The Monetary Transmission Process Recent Developme

The Monetary Transmission
Process
Recent Developments and Lessons for Europe
Edited by
The Deutsche Bundesbank
The Monetary Transmission Process
This page intentionally left blank

The Monetary Transmission
Process
Recent Developments and Lessons for
Europe
Edited by
The Deutsche Bundesbank

Editorial matter and selection © Deutsche Bundesbank 2001
Chapters 1–8 (and Comments/Discussions) © Palgrave Publishers Ltd 2001
All rights reserved. No reproduction, copy or transmission of

this publication may be made without written permission.
No paragraph of this publication may be reproduced, copied or
transmitted save with written permission or in accordance with
the provisions of the Copyright, Designs and Patents Act 1988,
or under the terms of any licence permitting limited copying
issued by the Copyright Licensing Agency, 90 Tottenham Court
Road, London W1P 0LP.
Any person who does any unauthorised act in relation to this
publication may be liable to criminal prosecution and civil
claims for damages.
The authors have asserted their rights to be identified
as the authors of this work in accordance with the
Copyright, Designs and Patents Act 1988.
First published 2001 by
PALGRAVE
Houndmills, Basingstoke, Hampshire RG21 6XS and
175 Fifth Avenue, New York, N. Y. 10010
Companies and representatives throughout the world
PALGRAVE is the new global academic imprint of
St. Martin’s Press LLC Scholarly and Reference Division and
Palgrave Publishers Ltd (formerly Macmillan Press Ltd).
ISBN 0–333–77244–X
This book is printed on paper suitable for recycling and

made from fully managed and sustained forest sources.
A catalogue record for this book is available
from the British Library.
Library of Congress Cataloging-in-Publication Data
The monetary transmission process : recent developments and
lessons for Europe / edited by The Deutsche Bundesbank.
p. cm.
Includes bibliographical references and index.
ISBN 0–333–77244–X (cloth)
1. Monetary policy—Europe. 2. Finance—Europe. I. Deutsche
Bundesbank.
HG925 .M6646 2000
332.4'94—dc21
00–042080
10 9 8 7 6 5 4 3 2 1
10 09 08 07 06 05 04 03 02 01
Printed and bound in Great Britain by

Antony Rowe Ltd, Chippenham, Wiltshire
Contents
Notes on the Contributors vii

Introduction
Heinz Herrmann 1
1 Analysis of the Monetary Transmission Mechanism:

Methodological Issues
Bennett T. McCallum 11
Discussion
Lawrence J. Christiano 44
Discussion
Harald Uhlig 51
2 Price Stability as a Target for Monetary Policy: De®ning and
Maintaining Price Stability
Lars E. O. Svensson 60
Discussion
Mervyn King 103
Discussion
~
JoseÂ Vinals 107
3 The Transmission Process
Allan H. Meltzer 112
4 Asymmetric Interest Rate Policy in Europe: Causes and
Consequences
Axel A. Weber 131
Discussion
Carlo A. Favero 159
Discussion
Philippe Moutot 164
5 Legal Structure, Financial Structure and the Monetary Policy
Transmission Mechanism
Stephen G. Cecchetti 170
Discussion
Manfred J. M. Neumann 195
v
vi Contents
Discussion
Ignazio Angeloni 201
6 Differences Between Financial Systems in European Countries:
Consequences for EMU
Reinhard H. Schmidt 208
Discussion
Charles A. E. Goodhart 241
Comment
Alain Vienney 247
7 European Labour Markets and the Euro: How Much
Flexibility Do We Really Need?
Michael C. Burda 252
Comment
Hendrik J. Brouwer 276
Comment
LuõÂs Campos e Cunha 279
8 The Monetary Transmission Process: Concluding Remarks
Otmar Issing 283
Francesco Giavazzi 294
Claes Berg 298
Ignazio Visco 303
Manfred J. M. Neumann 311
Name Index 315
Subject Index 319

Notes on the Contributors
lgnazio Angeloni, European Central Bank

Claes Berg, Sveriges Riksbank
Hendrik J. Brouwer, De Nederlandsche Bank NV
Michael C. Burda, Humboldt University, Berlin
LuõÂs Campos e Cunha, Banco de Portugal
Stephen G. Cecchetti, Federal Reserve Bank of New York
Lawrence J. Christiano, Northwestern University
Carlo A. Favero, Bocconi University
Francesco Giavazzi, Bocconi University
Charles A. E. Goodhart, London School of Economics and Political Science
Heinz Herrmann, Deutsche Bundesbank
Otmar Issing, European Central Bank
Mervyn King, Bank of England
Bennett T. McCalIum, Carnegie Mellon University
Allan H. Meltzer, Carnegie Mellon University
Philippe Moutot, European Central Bank
Manfred J. M. Neumann, University of Bonn
Reinhard H. Schmidt, Johann Wolfgang Goethe University, Frankfurt
Lars E. O. Svensson, Stockholm University
Harald Uhlig, Tilburg University
Alain Vienney, Banque de France
~
JoseÂ Vinals, Banco de Espana
lgnazio Visco, Organisation for Economic Cooperation and Development
Axel A. Weber, Johann Wolfgang Goethe University, Frankfurt
vii
Introduction
Heinz Herrmann
The debate on the monetary transmission process has not only continued in
the past few years but has also intensi®ed. There are a number of reasons for
this. In academic circles there are still various different approaches that seek to
explain how monetary policy in¯uences an economy. More and more areas
have come to be included in the analysis which attracted less attention in the
past. Thus, nowadays ± on the basis of an improved microeconomic theory ±
more attention is being paid to the role of ®nancial intermediaries and
®nancial systems in general than was previously the case.
In economic policy, central banks are confronted with the task of
safeguarding a level of price stability which, in most cases, is much greater
than that achieved in previous decades. How this target can best be achieved
is a question which depends in part on our state of knowledge of the
transmission process. In Europe, the debate has been given an additional
stimulus by the transition to European monetary union. How will this
changeover to a single currency in¯uence the effects of monetary policy in the
eleven member countries? Will the single monetary policy have similar
transmission effects in the different national economies or will marked
differences continue to exist? These are questions to which we are seeking to
®nd answers at the start of monetary union.
The Bundesbank therefore decided to invite economists from universities,
central banks and other institutions to a conference dedicated to these issues
on 26 and 27 March 1999. This volume contains the papers given and the
comments on them, with the introductory statements to a panel discussion
which concluded the conference (pp. 283±314).
Diverse models and methods have been developed in recent years aimed at
gaining a better understanding of the monetary transmission process. The
impression has arisen that the conclusions were not independent of the
methods of analysis used. For example, a study by the BIS (1995) found that,
depending on the type of model used, very different answers were provided to
the question of whether and how the transmission processes differ between
the countries examined.
1
2 Heinz Herrmann
Against this background Bennett McCallum in Chapter 1 considers

methodological aspects in Ànalysis of the Monetary Transmission Mechan-
ism: Methodological Issues', designed to help ®nd appropriate procedures and
models which allow meaningful statements to be made about how monetary
policy works through to the economy and about what constitutes a
meaningful monetary policy. McCallum argues that we should concentrate
on the systematic components of monetary policy and not on the impact of
monetary shocks. It follows that it is important to base the analysis on
structural models (rather than on vector-autoregression (VAR) models). To be
able to assess models, it is important to examine to what extent they are able to
re¯ect properties of the analysed economic variables as they act in reality. In
this connection McCallum stresses the importance of vector autocorrelation
functions as diagnostic tools, whereas he is critical of the much-used impulse
response functions for the respective policy shocks. He presents his preferred
approach using a sticky-price model. To demonstrate the effects of monetary
policy, he applies different interest rate rules, which may be regarded as
variants of a Taylor rule. Finally he discusses two other analytical approaches,
the narrative approach of Romer and Romer (1989) and a VAR procedure of
Sims (1998) in which the actual monetary policy stance in a historical period is
replaced by an alternative policy rule. But he is rather sceptical about both
approaches.
In his commentary, Lawrence Christiano considers both the methodologi-
cal proposals made by McCallum as well as the model used by McCallum by
way of illustration. He ®rst defends the approach of closely analysing the
impact of monetary shocks as an important method of selecting suitable
models and also mentions various new methods of testing the models'
consistency with reality. In this context, he also explains another method
which tests whether a model is capable of reproducing key features of
particularly interesting historical periods. He stresses the advantages of a
limited participation model of money over a sticky-price model which become
evident in the process. Harald Uhlig states that analysing both monetary
shocks and the systematic component of monetary policy are important and
that the two complement each other. He then asks some critical questions
about the model presented by McCallum. For example, he is of the opinion
that the exogeneity of investments in the model is unsatisfactory and that
some differences to which McCallum drew attention regarding the advantages
of certain monetary policy rules ultimately seem to have little relevance.
Finally, he makes some proposals as to how the model can be used
in the future for evaluating different monetary policy stances, given its
microeconomically-based structure.
The question of how monetary policy in¯uences real variables and prices
leads directly to considerations about how central banks should shape their
monetary policy. In Chapter 2, `Price Stability as a Target for Monetary Policy:
De®ning and Maintaining Price Stability', Lars Svensson addresses several
Introduction 3
problems: How is the general term `price stability' to be de®ned? How should
central banks formulate the rules which they follow? What are the lessons for
the European Central Bank (ECB)? Regarding the ®rst point, Svensson
discusses, inter alia, the advantages and disadvantages of in¯ation targets
versus price-level targets, the question of an appropriate price index for
formulating an in¯ation target and selecting the appropriate level for an
in¯ation target. Concerning the formulation of monetary policy rules,
Svensson draws attention to the advantages of forecast targeting over an
instrument rule and an intermediate targeting strategy. He then extends his
basic model to include non-linearities and stochastic coef®cients. He considers
a forecast targeting strategy which takes account of the probability distribu-
tion of the forecast to be an adequate response to such model extensions. He
advises the ECB to use a strategy of forecast targeting, to explicitly formulate
the in¯ation target, to publicise the in¯ation forecasts and to refrain from
assigning a special role to the money stock.
In his commentary, Mervyn King echoes Svensson's preferences for forecast
targeting, but warns against overemphasising the differences between the
strategies as applied in practice. He states it is more important that the
transmission mechanism is understood. He stresses the importance of looking
at the entire gamut of in¯ation forecasts when making monetary policy
decisions. Unlike Svensson, he considers it is justi®ed to assign a special role to
monetary growth in the monetary policy strategy. Jose Vin Ä als likewise sees
considerable advantages in a monetary policy strategy characterised as forecast
targeting by Svensson. His view is that such a strategy is consistent with major
principles of a good monetary policy practice, among which he includes
transparency and an orientation to the medium term. He expresses the view
that it is also much easier to communicate to the general public than an
optimal instrument rule.
Svensson has based his statements on monetary policy rules on a rather
generally formulated system of equations for describing the transmission
process in which a control variable, which is normally identical with a short-
term interest rate, stands in relation to predetermined variables (and possibly
forward-looking variables). In Chapter 3, `The Transmission Process', Allan
Meltzer looks more closely at this question, asking whether the importance of
money market rates or central bank rates is really that pivotal and exclusive in
the transmission process as is assumed in such models. Using examples from
US history, he seeks to show that focusing on a short-term interest rate is not
justi®ed. Rather, there had been periods in which monetary policy showed
effects which were not consistent with this usual conception of the monetary
transmission process. In his analysis he concentrates on historical phases of
de¯ation in which the development of the (real) monetary base explained
economic developments better than (real) short-term interest rates. Meltzer
mentions that two of these periods are suitable for showing that no liquidity
trap will emerge even when interest rates are low, a statement which takes on
4 Heinz Herrmann
added importance in the current phase of low interest rates. To support his
thesis of an effective wealth effect, he presents consumption functions in
which the real money stock has a clear explanatory value. He therefore
advocates keeping an eye on monetary growth when making monetary policy
decisions.
Empirical studies of the monetary transmission process which, on the one
hand, analyse the effects of the central banks' interest rate policy on the
economy and, on the other, ask which behaviour pattern is to be
recommended for monetary policy, are confronted with a severe methodo-
logical problem: it is not easy to decide whether the link between central bank
rates and/or money market rates and other economic variables such as
in¯ation is the result of a central bank's reaction function or whether it re¯ects
the implications of interest rate policy on the corresponding economic
variables. This prompts Axel Weber in Chapter 4, Àsymmetric Interest Rate
Policy in Europe: Causes and Consequences', to examine the link between
short-term interest rates and also in¯ation rates in the European countries (as
well as in the United States and Japan) and also the link between national
interest rates in Europe during the period of the European Monetary System
(EMS) and to demonstrate possible ways of better identifying the nature of the
links. Drawing on past papers (such as King and Watson, 1997), he uses
bivariate VAR models which take into account either interest rates and
in¯ation, or the German interest rate and that of another European country.
One of his ®ndings is that, in contrast to the United States, it is rather doubtful
whether German monetary policy can be accurately described by an interest
rate-smoothing rule if one assumes a long-run Fisher effect and long and
variable lags in monetary policy. Regarding the interest rate pattern in Europe,
he suggests that many central banks in Europe followed German interest rate
policy, but that there existed no simple long-run homogeneous pattern of
co-movement.
In his commentary, Carlo Favero agrees with the main thrust of Weber's
argument that considering monetary policy rules in isolation is a major
methodological problem. One should study such rules in the framework of
macromodels instead. However, he argues that bivariate models speci®ed in
®rst differences, as proposed by Weber, are only a ®rst step in the right
direction. It would be desirable to extend such a framework to achieve a closer
relation between theoretical and applied models. Philippe Moutot shares the
view that Weber raises an important question which should trigger a useful
research programme. Furthermore, he makes some more general remarks with
respect to monetary policy rules. He suggests that this chapter makes it clear
that it is dif®cult to identify the rules followed by the central banks under
consideration. Against this background, he warns against the hope that simple
rules can replace discretion. In his view the relevant issue is how to combine
the need for some limited discretion with the need for transparency vis-a-vis
the public, and describes how the ECB has proceeded in this respect.
Introduction 5
Differences in the monetary transmission process in European countries

may be caused by a variety of reasons. With increasing attention being paid to
the role assumed by information problems in the transmission mechanism ±
as is the case with the credit view, for example ± the focus has shifted to the
different organisation of ®nancial systems in the member countries. Both
Stephen Cecchetti and Reinhard Schmidt examine what in¯uence differences
in the ®nancial systems of the European countries may have on the
transmission of monetary policy.
Stephen Cecchetti in Chapter 5, `Legal Structure, Financial Structure
and the Monetary Policy Transmission Mechanism', examines the links
between the structure of ®nancial systems, the conditions for ownership rights
in a given country and the effectiveness of monetary policy. He bases his
considerations on the lending view, as presented, for example, by Kashyap and
Stein (1994). This implies that the central bank's interest rate policy should
have a stronger impact in countries in which enterprises mainly rely on bank
borrowing and in which the banking system is characterised by a multitude of
small and more fragile banks than in countries with robust capital market
®nancing and with dominant big banks. Against this background, Cecchetti
®rst identi®es marked differences in the ®nancial systems of the EMU member
countries. He then refers to comparative studies on the strength of the
transmission process in those countries which, in accordance with the lending
view theory, likewise show marked differences. In a next step Cecchetti looks
for the reasons that explain the different ®nancial structures. In this process he
focuses on the role of the respective legal systems, which result in different
rights of creditors and shareholders in the individual countries. He comes to
the conclusion that it is these legal differences which have led to the respective
®nancial structures and are thus one reason for the differences in transmission
mechanisms. He concludes that unless legal structures, are harmonised across
Europe, the ®nancial structures, and hence transmission mechanisms, will
remain diverse.
In his commentary, Manfred Neumann ®rst draws attention to some
inconsistencies in the construction of the indices which measure the
importance of the lending channel in the individual countries. He states that
the aggregation of different features to form such an index gives rise to dif®cult
and unresolved problems. Furthermore, he calls into question the robustness
of Cecchetti's empirical ®ndings on the links between the importance of the
credit channel in a given country and the impact of monetary policy on its
output. Ignazio Angeloni mentions, ®rst, that in most countries in the
monetary union the relevant legal systems are not so different. Secondly, he
expresses the view that the traditional transmission channels via the exchange
rate and the interest rates, which Cecchetti has neglected, are becoming more
and more similar in the wake of the changeover to monetary union. He also
puts forward the argument that the effects of a lending channel in the
European countries will converge as a result of forthcoming changes in the
6 Heinz Herrmann
respective ®nancial systems. Finally, he criticises some of Cecchetti's empirical

®ndings as contradicting the literature and also as being at variance with his
theoretical arguments.
Reinhard Schmidt, in Chapter 6, `Differences Between Financial Systems in
European Countries: Consequences for EMU' also comments on the effects
which typical differences in the ®nancial systems have on the transmission
mechanism and likewise focuses on the credit channel, in particular. He
describes some important features of a ®nancial system ± i.e. the ®nancial
sector and the ®nancial side of the real sector. He classi®es four sub-systems,
each of which may assume different forms and which, together, make up a
®nancial system: the ®nancial sector (bank-dominated versus ®nancial
market-dominated), the ®nancial patterns (bank lending versus ®nancing
via the capital market, bank deposits versus asset acquisition through
securities), corporate governance (control by insiders or outsiders) and the
corporate strategy (major versus minor importance of implicit contracts in
business decisions). He regards the German and the British systems as being
different but internally consistent, whereas he describes the French system as
being in a state of ¯ux in which it is dif®cult to ®nd a unifying `logique'. For the
purpose of assessing the monetary transmission mechanism, he differentiates
between the two levels ± between the central bank and the ®nancial sector, on
the one hand, and between the ®nancial sector and the real sector, on the
other. He argues that the central bank in the German system can exert a
greater in¯uence on the ®nancial sector than in the Anglo±Saxon system but
that in the former the ®nancial sector is much better able to cushion the
economy from monetary shocks. Although the differences at the respective
levels are considerable, they tend to offset each other in the aggregate. Finally
Schmidt comments on the stability of the ®nancial systems in the initial phase
of monetary union. The introduction of the Euro itself is unlikely to change
the systems fundamentally, but the increased interpenetration of elements
between one ®nancial system and another may have far-reaching implications
for the monetary transmission mechanism.
In his commentary, Charles Goodhart agrees with Schmidt's characterisa-
tion of the different ®nancial systems. But he is more critical regarding the
description of the transmission channels. He argues that the strength of the
credit channel depends very much on the given circumstances ± for example,
whether the banking system is well endowed with own capital. Regarding the
interest rate channel he believes that the monetary policy effects are
particularly strong in the United Kingdom. In this connection he raises the
question why UK interest rates have shown greater volatility than German
rates in the past. Finally, Goodhart notes that he is less pessimistic than
Schmidt regarding the coming restructuring of the ®nancial systems and the
associated problems. This view is shared by Alain Vienney. Vienney also
criticises the overemphasis on the credit channel compared with the interest
rate channel. In his view, neglecting the interest rate channel is the reason
Introduction 7
why Schmidt overstates the differences in the transmission of monetary policy

between the Continental Europen countries, whereas he understates the

differences vis-a-vis the Anglo±Saxon countries. In this context, Vienney also
draws attention to the differences in the maturities of ®nancial relationships
and in the balance sheet structures which still exist in the United Kingdom
compared with continental Europe.
A major factor in¯uencing the transmission of monetary policy, in addition
to the ®nancial systems and the situtation on goods markets, are labour
markets. A key factor here is the ¯exibility of prices and wages. Two concerns
have emerged with a view to EMU: on the one hand, the fear that in a single
currency area there will be a convergence of wages which is not accompanied
by a corresponding convergence of labour productivity. The second concern
in this respect is that, with a single currency and a single monetary policy,
there will be no mechanisms to deal with asymmetric shocks. Hence, the
question arises of whether there are indications that the labour markets will
show suf®cient ¯exibility to assume the necessary burden of adjustment. In
Chapter 7, Èuropean Labour Markets and the Euro: How Much Flexibility Do
We Really Need?', Michael Burda attempts to provide answers to that question.
Burda ®rst discusses the general question of what role rigidities on the labour
and goods markets may play in the convergence of countries in a monetary
union and summarises his ®ndings by saying that, at the start of monetary
union, rigidities and the lack of mobility on the labour markets must be
considered problematical. He then outlines scenarios which show the future in
a much more favourable light. He takes the view that monetary union will act
as the engine for the freeing-up of labour markets and real wages in Europe. A
factor contributing to this, he suggests, is that prices will tend to adjust more
slowly to a new equilibrium since, inter alia, he considers the adjustment
pressures in a large, more closed currency area to be lower in this respect. On
the other hand, he envisages that the negotiating power of the trade unions
will decline in the future Europe and that the pressure on governments to carry
out institutional reforms ± for example, of the social security systems ± will
increase, which in turn will boost the ¯exibility of wages. In view of this new
combination of circumstances, Burda predicts an improved transmission of
monetary policy, but he also warns against abuse.
In his commentary, Hendrik Brouwer stresses how important it is for labour
markets to become more ¯exible in view of the high level of unemployment
and potential asymmetric shocks in monetary union. Although he supports
some of Burda's arguments ± for example, that goods prices might become less
¯exible ± he disagrees with him in some important points. He mentions that
there are indications of persisting labour market rigidities and he is less
optimistic about the governments' commitment to reforms. Luis Campos e
Cunha, too, is sceptical regarding Burda's central statement that the euro will
be the `Trojan horse' which liberalised labour markets. In addition, he is not
convinced by Burda's opinion that EMU will lead to less ¯exible prices. Finally,
8 Heinz Herrmann
he warns of the risk of inappropriate harmonisations of labour markets ± for

example, of minimum wages.
The conference volume ends with statements during a debate which was
intended to re¯ect, in particular, the view of practitioners on the topics
discussed and the problems ahead. In his introductory remarks Otmar Issing
addresses two questions, in particular: `What do we know about the
transmission process in Europe at the start of monetary union?' and `How
will this process change in the coming years?' He reminds us that the
participating countries share a number of common features, but also have
some differences. But in his view, it would be dif®cult to make any clear-cut
pronouncements, based on institutional differences or econometric studies,
on the divergent effects of monetary policy in those countries. Looking to the
future, he is of the opinion that a number of factors, not least the single
monetary policy and the single currency, should result in a convergence of the
transmission process. However, he draws attention to our imperfect knowl-
edge, a factor which the ECB's monetary strategy takes into account. Francesco
Giavazzi focuses in particular on the implications of the restructuring of the
European banking markets. He refers to studies according to which the credit
channel works differently in the individual countries. It seems likely that
cross-border banking mergers would reduce such differences; actually,
however, most mergers in Europe take place within national boundaries.
Giavazzi states that this is worrying, for competition reasons, on the grounds
of general considerations concerning an ef®cient ®nancial system in Europe
and also in the context of a single transmission process. Claes Berg ®rst
presents his view on the previous debate and questions ± in particular, Burda's
optimistic conclusions regarding the ¯exibility of real wages. In his more
general remarks he comments on the monetary policy strategy, emphasising
the importance of published in¯ation forecasts. Futhermore, he considers
whether a central bank should pursue an aggressive interest rate policy or
whether a step-by-step policy would be preferable. He mentions that
uncertainty does not always justify a smooth interest rate policy. Based on
the economic situation in spring 1999, Ignazio Visco discusses some future
challenges for economic policy and, in particular, for monetary policy. He
stresses the progress achieved in convergence and in economic integration in
Europe which, in principle, should facilitate a European monetary policy. At
the same time he points to the uncertainty as to the appropriate monetary
policy in the new currency area, an uncertainty engendered in part against the
backgound of the historically low in¯ation rates. In this connection he also
emphasises the necessity of a discussion on the appropriate policy mix in
Europe. Manfred Neumann discusses two questions: `What is the optimal
®nancial system?' and Àre differences in the ®nancial systems of importance
at all in the long run?' He comes to no de®nite conclusion regarding the
advantages of the respective ®nancial systems. He stresses, however, that a
system which is based more strongly on ®nancial markets (rather than on
Introduction 9
banks) is likely to have some advantages for the funding of innovative, young
enterprises. At the same time, he states that a comparison of the growth
performance of the United States and that of Germany does not indicate the
superiority of the US ®nancial system. Also, he takes the view that differences
between countries are a problem, especially whenever unexpected shocks
emanate from monetary policy. From this he concludes that monetary policy
should have a medium-term orientation as far as possible in order to avoid
such shocks.
References
BIS (1995) `Financial Structure and the Monetary Policy Transmission Mechanism',
Basle.
Kashyap, A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending,' in
N. G. Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press.
King, R. G. and M. W. Watson (1997) `Testing Long Run Neutrality', Federal Reserve Bank
of Richmond, Economic Quarterly, 83, pp. 69±101.
Romer, C. and D. Romer (1989) `Does Monetary Policy Matter? A New Test in the Spirit
of Friedman and Schwartz', NBER Macroeconomics Annual 1989, Cambridge, MA, MIT
Press.
Sims, C. A. (1998) `The Role of Interest Rate Policy in the Generation and Propagation of
Business Cycles: What has Changed Since the '30s?', in Federal Reserve Bank of Boston
(ed.), Beyond Shocks: What Causes Business Cycles?, Boston: Federal Reserve Bank of
Boston.
1
Analysis of the Monetary Transmission

Mechanism: Methodological Issues
Bennett T. McCallum1
1 Introduction
The purpose of this chapter is to consider several methodological issues

relevant for study of the monetary transmission process. These issues involve
relative emphasis on monetary shocks as opposed to systematic policy
adjustments; vector autoregression versus structural modelling research
strategies; impulse response versus vector autocorrelation functions as
diagnostic tools; and an evaluation of the so-called `narrative approach'. But
while these methodological issues are stressed, the chapter's approach is
signi®cantly substantive, in the sense that the issues will be considered in the
context of a non-trivial quantitative analysis that is intended to be of interest
on its own.
As a preliminary matter, it may be useful to outline what meaning is here
being given to the term `monetary transmission mechanism'. That this term
evokes different responses from different scholars is well illustrated by a
symposium on `The Monetary Transmission Mechanism' featured in the Fall
1995 issue of the Journal of Economic Perspectives. In the papers of that
symposium, Bernanke and Gertler (1995) focus on the credit channel; Meltzer
(1995) promotes monetarist emphasis on the importance of recognising
multiple assets;2 Taylor (1995) outlines a particular econometric framework
for studying the transmission mechanism; Obstfeld and Rogoff (1995) discuss
foreign exchange rate policy and ®nancial crises; and Mishkin (1995) provides
a brief overview. More generally, many writers on the subject restrict their
attention to the effects of monetary policy shocks,3 while some are concerned
only with effects on real variables. In the present chapter, however, the
concept of the transmission process to be considered includes effects on both
real and nominal variables of shocks and more especially the regular,
systematic component of monetary policy. The implied de®nition, therefore,
is similar to that expressed by Taylor (1995, p. 11): `the process through which
monetary policy decisions are transmitted into changes in real GDP and
in¯ation.'
11
12 Bennett T. McCallum
An outline of the chapter is as follows. In Section 2, it is argued that study of

the systematic component of monetary policy actions is at least as important
as the study of the unsystematic component, also known as policy shocks.
Section 3 then presents some procedures for exhibiting effects on in¯ation,
output and other variables of different systematic policy responses. These
differences are, of course, model-speci®c: they depend upon the structural
speci®cation of the model being utilised. In Section 4, consequently, some
variants of the basic model utilised in Section 3 are considered. It is
demonstrated that systematic policy effects are signi®cantly dependent upon
speci®cations relating to price adjustment behaviour, habit formation in
saving versus consumption decisions and the economy's openness to foreign
trade. This dependence is expressed in terms of root-mean-square statistics for
in¯ation targeting errors and output gap measures, and also in terms of
the characteristics of impulse response functions for shocks other than the
monetary policy shock. Section 5 concerns diagnostic tools to be used in the
construction of structural models; here it is suggested that more attention
should be given to vector autocorrelation functions (and correspondingly less
to impulse response functions) than is typically the case in the vector-
autoregression (VAR) literature. Finally, Section 6 offers a partial evaluation
and criticism of the `narrative approach' introduced by Romer and Romer
(1989) and a non-standard VAR procedure utilised by Sims (1998b), plus brief
comments on relevant papers by Bernanke, Gertler and Watson (1997) and
Dotsey (1999). Section 7 is a brief conclusion.
2 Shocks versus systematic policy
A large volume of literature exists, much of it highly sophisticated, in which

the effects of monetary policy on output, prices, and other variables are
discussed entirely in terms of policy shocks.4 In this context, policy shocks
represent the random, unsystematic component of the monetary authorities'
actions ± i.e. the portion that is not related to the state of the economy, current
or past. A leading theme of the present chapter is that emphasis on the shock
component has been overdone; that while both shocks and the systematic
component of behaviour are important, it would be more fruitful to emphasise
the latter. This point of view has been taken by a number of analysts, including
Taylor (1995), Rotemberg and Woodford (1997, 1999) and Bernanke, Gertler,
and Watson (1997), but needs to be stressed nevertheless because of the sheer
volume of literature that differs in this crucial respect.
Perhaps the simplest way of arguing for an emphasis on the systematic
component of policy is to recognise that quantitatively the unsystematic
portion of policy instrument variability is quite small in relation to the
variability of the systematic component. An illustration is provided by the
prominent study by Clarida, Gali and Gertler (1998) of policy behaviour since
1979 by central banks of the G-3 nations. In particular, their `baseline'
Methodological Issues 13
estimations of monthly Bundesbank, Bank of Japan and Federal Reserve

reaction functions indicate that the fraction of monthly interest instrument
variability that is unexplained by systematic determinants is only 1.9, 3.0 and
1.6 per cent, respectively.5 Also, Rotemberg and Woodford (1997) and
McCallum and Nelson (1998) ®nd in the US quarterly data that only about
5 per cent of instrument variability is unexplained over roughly the same
period.
Indeed, it is conceivable that policy behaviour could be virtually devoid of
any unsystematic component. In the limit, that is, the variance of the shock
component could approach zero. But this would not imply that monetary
policy is unimportant for price-level behaviour, a central bank's main
responsibility. Nor would it imply that policy is unimportant for real cyclical
activity unless the economy is of the rather special type that satis®es the
`policy ineffectiveness' proposition.6 More generally, it should be kept in mind
that when a central bank raises its interest rate instrument by (for example) 50
basis points ìn order to head off in¯ation', the action is likely to represent a
systematic response, not a shock.
To illustrate the implications of some less extreme and less obvious
phenomena, let us consider a simpli®ed analytical representation of monetary
policy behaviour and its consequences for output and in¯ation. Here (and in
the remainder of the chapter) let Rt , yt and pt denote a short-term nominal
interest rate, the logarithm of real output, and the log of the price level. Also let
yt be the natural rate value of yt , with pt the associated price level,7 and let vt
and et represent shocks to spending and monetary policy behaviour. Thus et is
the unsystematic component of policy. To keep the present example simple,
we temporarily pretend that yt is a constant and normalise it as y 0, so that
yt also measures output relative to its natural rate value. Our schematic model
is given by three equations, as follows:
yt b0 b1 Rt � Et pt1 Et yt1 vt b1 < 0 1:1

pt � pt�1 1 � p t�1 � pt�1 Et�1 p t � p t�1 0 < < 1 1:2

Rt Et�1 pt1 0 1 Et�1 pt1 � et 1 > 0 1:3
The ®rst of these is an IS-type relation representing demand for current output
± i.e. saving versus spending behaviour. Optimising theory suggests that the
variable Et yt1 should appear as indicated on the right-hand side; this has been
argued by Kerr and King (1996), McCallum and Nelson (1999), and Rotemberg
and Woodford (1997), among others. Neglect of this term would not simplify
derivations substantially and would not affect the main points of the
analysis.8 This future expected income term will accordingly be incorporated
in the quantitative models of Sections 3±4.
Equation (1.2) is one form of the P-bar price adjustment relation that was
rationalised and utilised in papers by McCallum and Nelson (1997, 1998). It is
not as widely used as the Calvo±Rotemberg (1983) model9 or variants of the
Fuhrer±Moore (1995) speci®cation, but its theoretical properties are arguably

superior.10 In any event, most of the points to be made here would carry over
to other price adjustment speci®cations. And in our quantitative work below,
the Fuhrer±Moore speci®cation will be considered in addition to (1.2). Note
that the variable pt in (1.2) is the value of pt that would induce producers to
make yt y t .
Equation (1.3) represents monetary policy behaviour according to which an
interest rate instrument Rt is set each period so as to raise the expected real rate
of interest, Rt ± Et�1 pt1 , when the expected future in¯ation rate exceeds the
target value . With 0 chosen to equal the average real rate of interest, (1.3)
amounts to a special case of a forward-looking Taylor rule. Expectations are
dated t � 1 in (1.3) so as to realistically limit the information available to the
central bank when setting Rt . The notation is that Et ztj Eztj j
t , where
t
includes variables dated t and earlier.
Solution of the foregoing model is facilitated by the fact, demonstrated in
McCallum and Nelson (1998), that (1.2) implies that Et�1 y~ t y~ t�1 , where
y~ t yt ± yt . In the present setting with yt 0, yt y~ t so we have Et�1 yt yt�1 .
Then the minimal-state-variable (MSV) rational expectations solution11 for yt
and pt is of the form
yt 10 11 yt�1 12 vt 13 et 1:4

pt 20 21 yt�1 22 vt 23 et 1:5
and it is clear that 10 0, 11 . Also, since prices are fully predetermined,
22 23 0. Given these facts, the undetermined coef®cient procedure can
be used to ®nd the remaining values of the ij . The solution, it turns out, is
yt yt�1 1 =1 1 1 � vt b1 1 =1 1 1 � et 1:6

pt � 0 =1 � b0 =b1 1 1 � =b1 1 yt�1 1:7
Now let us suppose that the above system were studied by some VAR approach
that correctly identi®ed the policy shock term et . The coef®cient
b1 1 /(1 1 )(1 ± ) is negative, so a VAR estimate of (1.6) would correctly ®nd
a negative effect from a positive shock to Rt , provided that there had been
enough sample-period variation in et . If et had stayed close to its mean value of
zero, however, statistical procedures might ®nd no signi®cant effect in ®nite
samples of a realistic size.
In the case of pt , (1.7) indicates that the policy shock et plays no role when
the effect of yt�1 is taken into account, as it would be in any VAR study. A study
that looked for monetary effects on in¯ation by the Granger-causality method
would therefore conclude that there are no such effects. The method relying
upon the `fraction of explained variance' of the various shocks would attribute
some portion to et since et�1 , et�2 , . . . help to explain yt�1 . But the fraction of
pt variability attributed to monetary shocks would then be precisely the same
as the fraction pertaining to yt variability. Note that since (1 ± )/b1 1 < 0,
these will be such that a surprise increase in Rt will have the effect of increasing
± rather than decreasing ± subsequent values of pt . This is a rather perverse
property of this simpli®ed model.12 But it is nevertheless true that an increased
value of the policy response parameter 1 will decrease the variability of
in¯ation, pt; in the system (1.1)±(1.3).
Furthermore, note that, since the unconditional mean Eyt equals zero by
construction, relation (1.1) implies that the average value of the real rate of
interest rt Rt ± Et pt1 equals ±b0 /b1 . Thus if the central bank chooses 0 to
equal ± b0 /b1 , as a sensible policymaker would do, (1.7) reduces to
pt 1 � =b1 1 yt�1 1:7
Thus on average, over a large number of periods, realised in¯ation will tend to
equal , a target rate that is entirely determined by monetary policy.13 Within
a given policy regime, the `long-run' ± i.e. unconditional mean ± value of
in¯ation is entirely monetarily determined, but this fact would not be revealed
by shock-oriented VAR procedures.14
Having argued that it is more important to focus on the systematic portion
of monetary policy actions, rather than on shocks, we are then necessarily
driven toward the study of structural models, rather than VARs.15 The reason is
that even ìdenti®ed' or `semi-structural' VAR systems do not give rise to
behavioural equations that can be presumed to be structural ± i.e. policy-
invariant. The purpose of identi®ed VARs is to identify the unsystematic
component of monetary policy, not to generate policy-invariant equation
systems. But it is the latter that governs the effects of systematic or anticipated
policy actions. This should be emphasised, for it is the crucial point of my
argument. It is, I believe, consistent with the analysis and views of most
creators and practitioners of the identi®ed VAR approach, including Bernanke,
Blanchard and Watson, and Christiano, Eichenbaum, and Evans.16 My
position would not be accepted by Sims (1998b), however, so some discussion
of his approach is included below (in Section 6). It should be noted that my
objections to several aspects of VAR analysis are not the same as those put
forth by Rudebusch (1998). In fact, they are not actually objections to VAR
analysis per se but rather are arguments for concentrating on the systematic
component of policy rather than the shocks.
3 Effects of systematic policy
In this section the purpose is to describe one approach to analysis of the effects
of the systematic component of monetary policy. Since this undertaking
requires use of a structural model, according to the foregoing argument, the
results obtained will depend upon the adopted speci®cation of economic
behaviour. My starting point will be the small-scale, open economy, quarterly
model based on an optimising analysis that is developed and presented in
McCallum and Nelson (1998). The following paragraphs will brie¯y outline
that model and report some simulation results that serve to characterise the
effects of monetary policy. Variants of the basic model will be considered in
Section 4.
Basing one's analysis on the assumption of explicit optimising behaviour by
the modelled individuals in a general equilibrium setting is obviously not
suf®cient ± and perhaps not necessary ± for the creation of a structural model
that is speci®ed with reasonable accuracy relative to economic reality. The
optimising general equilibrium approach can be helpful in this respect,
however, since it eliminates potential internal logical inconsistencies that are
possible when this source of intellectual discipline is absent. The model in
McCallum and Nelson (1998), henceforth termed the M±N model, has a very
simple basic structure since it depicts an economy in which all individuals are
in®nite-lived and alike. As with many recent models designed for policy
analysis, it assumes that goods prices are `sticky,' i.e. adjust only slowly in
response to changes in conditions. It differs from many previous efforts in this
genre, however, in three ways. First, the gradual price-adjustment speci®cation
satis®es the strict version of the natural-rate hypothesis.17 Second, the
modelled economy is open to international trade of goods and securities.
And, third, individuals' utility functions do not feature time-separability, but
instead depart in a manner that re¯ects `habit formation'.
This last feature is speci®ed as follows. A typical agent desires at t to
maximize Et Ut Ut1 . . ., where the within-period measure Ut is speci®ed
as
Ut expvt = � 1Ct =Ct�1 h �1= 1 � �1 Mt =pt 1� 1:8
Here Ct is a CES consumption index, Mt /Pt is real domestic money balances, vt

is a stochastic preference shock and h is a parameter satisfying 0 h < 1. With h
0, preferences feature intertemporal separability, but with h > 0 there exists
`habit formation' that makes consumption demand less volatile.
The open-economy aspect of the model is one in which produced goods
may be consumed in the home economy or sold abroad. Imports are
exclusively raw materials, used as inputs in a production process that
combines these materials and labour according to a CES production function.
Capital accumulation is not modelled endogenously, but securities are traded
internationally. The relative price of imports in terms of domestic goods, i.e,
the real exchange rate, affects the demand for exports and imports, the latter
in an explicit maximising fashion. Nominal exchange rates and the home
country one-period nominal interest rate are related by a version of uncovered
interest parity, one that realistically includes a stochastic `risk premium' term
(as in Taylor, 1993b, and many multi-country econometric models).
Price adjustments conform to the P-bar model, mentioned above, but with
capacity output y t now treated as a variable that depends upon raw material
inputs and the state of technology, the latter driven by an exogenous

stochastic shock that enters production in a labour-augmenting fashion.18 As
mentioned above, price adjustment behaviour implies Et�1 y~ t y~ t�1 , so
application of the unconditional expectation operator yields Ey~ t Ey~ t and
this implies Ey~ t 0 regardless of the monetary policy rule employed. This
natural rate property is not a feature of the Calvo±Rotemberg or Fuhrer±
Moore models of price adjustment. Indeed, there are very few sticky-price
models that have the natural rate property, the only other one that I know of
being Gray±Fischer-style nominal contracts that imply limited persistence of y~ t
magnitudes.
The foregoing is intended to give the reader a broad overview of the M±N
model; for a full description the reader may consult McCallum and Nelson
(1998).19 Here the objective is to combine that structure with a policy rule
formulation that permits a moderately straightforward `measure' of the effect
of systematic policy activism. We begin with the following rule:
Rt 1 � 3 Et�1 pt1 0 1 Et�1 pt �

2 Et�1 y~ t 3 Rt�1 et 1:9
In (1.9), one difference relative to (1.3) is the inclusion of a lagged Rt term, to

re¯ect a form of interest rate smoothing that seems to characterise the
behaviour of actual central banks. In light of estimates and previous
experience, our basic experiments will assign a value of 0.8 for 3 . A second
difference is the appearance of 2 Et�1 y~ t , as in the Taylor rule (Taylor, 1993a),
but initially we shall set 2 0. Then with 2 0 and 3 0.8, the extent to
which activist but systematic policy actions are taken is directly related to the
magnitude of 1 . In that regard, a third difference relative to (1.3) is that
feedback is taken from Et�1 pt , rather than Et�1 pt1 . The reason is that the
former is more effective in this model, as will be seen below. Since the implicit
primary objective of rule (1.9) is to keep in¯ation pt close to the target value
, one measure of the effect of policy is the reduction (if any) in the root-
mean-square-error (RMSE) value of pt � . In the simulations reported in
this section, all constant terms are set to zero ± a standard practice in stochastic
simulation work of this type20 ± so the standard deviation of pt can be
interpreted as the RMSE value of pt � . Somewhat less tenuously, the
standard deviation of y~ t can be interpreted as the RMSE value of yt ± yt . In all
cases, the reported magnitudes are mean values averaged over 100 replica-
tions, with each simulation pertaining to a sample period of 200 quarters (after
53 start-up periods are discarded). Calculation of the rational expectation
solutions is effected by means of Klein's (1997) algorithm.
The most basic results are given in Table 1.1. In the ®rst row of the ®rst panel
we see how the RMSE value of p � decreases as additional policy response
to expected target errors is applied. With 1 0.1, there is almost no response
to such errors; with 1 0 there would merely be a gradual adjustment of
Table 1.1 Simulation results for variants of policy rule (1.9)

Basic model. 3 0.8
Case std dev. of 1 0.1 0.5 1.0 5.0 50.0
Et�1 pt as target p 10.54 7.37 5.48 1.91 0.24

2 0 y~ 1.45 1.83 2.12 2.71 3.08
R 2.17 2.55 2.86 3.77 4.50
Et�1 pt1 as target p 11.21 9.76 8.57 4.51 **

2 0 y~ 1.35 1.52 1.65 2.29 **
R 2.06 2.14 2.29 3.25 **
2 0.1 0.5 1.0 5.0 50.0

1 0.5 on pt p 7.45 8.29 9.10 13.37 18.77
2 on Et�1 y~ t y~ 1.79 1.70 1.59 0.97 0.17
R 2.54 2.96 3.32 4.97 6.98
Notes: The 2 values actually used are 1 /4 of the values listed; the latter correspond to units of
measurement in annualised percentage points, as are typically reported in the Taylor rule literature.
That statement applies to all results in this chapter. means MSV solutions are not available with
existing software.
Rt � Et�1 pt1 toward its long-run average value. As 1 is increased, with policy
response strength increased, the standard deviation of pt falls distinctly.
In the second panel, feedback response is taken from Et�1 pt1 � .
Clearly, the variability of pt is much greater than when Et�1 pt is the target
variable, especially for large values of 1 . In the model at hand, then, the
stabilising effect of monetary policy on the in¯ation rate is greater when
Et�1 pt , rather than Et�1 pt1 , is the variable responded to. That property
does not obtain for all model speci®cations, of course.
In both of the ®rst two panels, we see that application of stronger feedback to
in¯ation rate discrepancies has the effect of increasing the variability of y~ t , the
output gap. In the third panel we consider policy responses to the output gap, as
well as to in¯ation. In particular, we assume that the interest rate instrument is
adjusted upward when Et�1 y~ t is positive ± i.e. when output is expected to exceed
its natural rate value. As 2 is increased ± i.e. moving to the right in the table ± we
see that the variability of y~ t falls, as one would expect. Thus it is the case that
systematic monetary policy can be used to stabilise output (i.e. keep yt close to y t )
in this model, despite its highly classical long-run properties. When 2 is
increased with constant 1 and 3 , the variability of in¯ation increases.
Another way to see the effect of the systematic portion of monetary policy is
to compare impulse response functions for different values of policy rule
parameters. Let us return to the case represented in the ®rst panel of Table 1.1
i.e. with 2 0, 3 0.8, and 1 varied over a wide range. In this context, the
impulse response function for the policy shock itself is not as interesting as for
some of the other shocks. Let us consider ®rst a shock to the expectational IS
function ± i.e. a shock to preferences that increases the demand for current
consumption in relation to future consumption. Impulse response functions

for the variables yt , pt , pt , qt , st , and Rt are shown in Figure 1.1a (for a unit
shock) when 1 0.1 ± i.e. when policy response is very weak. By comparison,
the same responses are shown in Figure 1.1b for a very strong policy response,
with 1 5. We see that the response of output to the shock is slightly greater
in the second case, but that the responses of in¯ation and the price level (pt
and pt ) are much smaller; please note the vertical axis scalings. Furthermore,
the response pro®les for qt and st , the real and nominal exchange rates, are not
even of the same shape in the two panels.21 Clearly, the systematic
component of monetary policy has major effects on the way in which the
economy responds to demand shocks (IS shocks) in this model.
Also of interest is the difference of the response patterns to a shock to the
uncovered interest parity (UIP) relation ± i.e. a foreign exchange risk premium
shock. Figure 1.2 includes panels for the same two policy rules as re¯ected in
Figure 1.1. Thus the top panel, Figure 2a, obtains when policy responds very
weakly to in¯ation target misses while the bottom panel, Figure 2b, is for very
strong responses. Again, the responses of pt and pt are distinctly muted by
stronger policy behaviour (larger 1 values). By the òvershooting' mechan-
ism, consequently, the nominal and real exchange rate responses are larger
when 1 is large.
Finally, let us consider a technology shock, one that increases the value of
yt .22 Output and real exchange rate responses are not dissimilar with 1 0.1
and 1 5.0, but the response of nominal variables is drastically different with
the different 1 values ± see Figures 1.3a and 1.3b. With 1 0.1, in¯ation falls
and then very slowly returns to zero, whereas with 1 5.0 in¯ation brie¯y
rises. As a result, the time pro®les for the price level and the nominal exchange
rate are extremely different. All in all, the differences depicted in Figures 1.1±
1.3 re¯ect the effects of the systematic component of monetary policy
behaviour in response to shocks of the type that are crucial in the
implementation of monetary policy.
4 Model speci®cation
The previous section has suggested some procedures for characterising the
effects of the systematic component of monetary policy for a given structural
model. But of course different models generate very different effects, so it is
essential to have a strategy for developing a good structural model. Most
researchers would agree that it is desirable for a model to be consistent with
both economic theory and empirical evidence, but that dual requirement is
only a starting point for consideration of numerous issues.
Like many economists, I have been persuaded that it is a desirable practice to
begin with the construction of a general equilibrium model in which
individual agents are depicted as solving dynamic optimisation problems. As
mentioned above, such a step is neither necessary nor suf®cient for obtaining
0.2 0.2
0.1 0
y
q
0 –0.2
0 10 20 0 10 20
0 0
p
s
–0.5 –0.5
0 10 20 0 10 20
0 0.05
–0.1 0
R
Δp
–0.2 –0.05
0 10 20 0 10 20
a Responses to unit shock to IS; μ1 =

= 0.1
0.1
0.2 0.4
0.1 0.2
q
y
0 0
0 10 20 0 10 20
0 0.5
0
S
p
–0.05 –0.05
0 10 20 0 10 20
0 0
Δp
–0.01 –0.02
R
–0.02 –0.04
0 10 20 0 10
20
b Responses to unit shock to IS; μ1 = 5.0
Figure 1.1 Impulse response functions, basic model (IS shock)

0.1 2
0.05 1
y
q
0
0
0 10 20
0 10 20
0
2
–0.5 0
p
s
–1 –2
0 10 20 0 10 20
0 0
–0.5 –0.05
Δp
–1 –0.1
0 10 20 0 10 20
a Response to unit shock to UIP; μ1 = 0.1
0.4 4
0.1 2
q
y
0 0
0 10 20 0 10 20
0 5
–0.1 0
s
p
–0.2 –5
0 10 20 0 10 20
0.2 0
0 –0.1
R
Δp
–0.2 –0.2
0 10 20 0 10 20
b Response to unit shock to UIP; μ1 = 5.0
Figure 1.2 Impulse response functions; basic model (UIP shock)

0.4 1
0.2 0.5
q
y
0 0
0 10 20 0 10 20
0 2
–1 0
s
p
–2 –2
0 10 20 0 10 20
0 0.1
Δp
–0.1 0
R
–0.2 –0.1
0 10 20 0 10 20
– μ = 0.1
a Responses to unit shock to y; 1
0.4 1
0.2 0.5
q
y
0 0
0 10 20 0 10 20
0.05 1
0 0.5
p
–0.05 0
0 10 20 0 10 20
0.05 0.1
Δp
0 0
R
–0.05 –0.1
0 10 20 0 10 20
– μ = 5.0
b Responses to unit shock to y; 1
Figure 1.3 Impulse response functions, basic model (y shock)

Table 1.2 Simulation results for model variants

2 0.0; 3 0.8
Case std. dev. of 1 0.1 0.5 1.0 5.0 50.0
h 0.8 p 10.54 7.37 5.48 1.91 0.24

Et�1 pt as target y~ 1.45 1.83 2.12 2.71 3.08
R 2.17 2.55 2.86 3.77 4.50
h 0.0 p 7.01 6.09 5.33 2.54 0.37

Et�1 pt as target y~ 1.38 2.02 2.83 5.47 7.11
2 0 R 3.35 4.02 5.03 8.09 9.88
Closed econ. p 18.33 5.98 2.87 0.59 0.06

Et�1 pt as target y~ 1.50 1.42 1.41 1.42 1.44
R 1.47 1.03 0.99 0.86 0.84
F±M price adj p 3.87 3.48 3.15 2.54 1.87

Et�1 pt as target y~ 1.77 2.14 2.45 3.96 8.27
R 2.84 3.15 3.51 6.87 27.56
F±M price adj p 3.97 3.07 3.36 2.63 2.00

Et�1 pt1 as target y~ 1.81 2.03 2.39 3.54 6.81
R 2.89 3.02 3.57 5.99 20.47
a good model, but is useful in tending to reduce inconsistencies and forcing

the modeller to think about the economy in a disciplined way. But adherence
to dynamic optimising general equilibrium analysis still leaves room for
enormous differences among models ± even ones that are of the same scale and
include the same variables. In this section I will attempt to discuss a few of the
crucial speci®cational issues, illustrating their importance by various compar-
isons with the model introduced in section 3.
One non-standard feature of that model is the presence of `habit formation'
in consumption behaviour. Much more common is a speci®cation with a
time-separable intertemporal utility function ± as utilised, for example, by
Rotemberg and Woodford (1997), Kerr and King (1996), or McCallum and
Nelson (1997). This more standard speci®cation can be obtained as a special
case of the model of Section 3 simply by setting the parameter h at the value of
zero ± see (1.8). If that is done and simulations like those of Table 1.1 are
conducted, the results with Et�1 pt as the target variable are shown in the
second panel of Table 1.2, where the ®rst panel repeats a comparable reference
case from Table 1.1. It will be seen that the differences in RMSE values for
in¯ation (for various 1 values) are somewhat smaller than in the reference
case, and that the RMSE differences for y~ t are substantially larger. Intuitively,
setting h 0.0 eliminates a source of consumption persistence that obtains

with h 0.8. This change leaves output free to respond more strongly to
shocks and simultaneously makes in¯ation more controllable by policy
actions.
The difference in behaviour when h 0 rather that h 0.8 shows up even
more dramatically in impulse response functions. For the purpose of this
comparison, we set 1 at the intermediate value of 1.0 while keeping 2 0
and 3 0.8. The responses to a monetary policy shock are shown in Figures
1.4a and 1.4b, where the case with h 0 is in part b. Comparing the upper left
panels from the two sets, we see that the response of yt to a unit Rt shock is
almost three times as large when habit formation is absent. Even more
drastically, we see that with h 0 the response of in¯ation (and the price level)
is in the opposite direction from that with h 0.8: an unanticipated increase
in Rt causes pt to rise temporarily, rather than fall. This is the same property
that was observed to obtain in the simple analytical model of Section 2. Most
readers will probably ®nd it implausible, although it is reminiscent of the
`price puzzle' empirical results that have some tendency to arise in VAR studies
with models that include no commodity-price variable (Sims, 1992). But the
main point to be stressed here is that behaviour is quite different in models
with h 0 and h 0.8.23 This difference also obtains in response to a vt shock,
as is illustrated in Figure 1.5.
Next we consider the effect of removing the open-economy features of our
basic model ± i.e. converting it to a closed-economy speci®cation.24
Comparing the third panel of Table 1.2 with the ®rst (reference) panel shows
that the in¯ation RMSE values are much more sensitive to 1 in the closed-
economy speci®cation ± i.e. in¯ation is more readily controllable by monetary
policy. The different 1 settings have much less effect on y~ t variability,
however. One problem with the open-economy speci®cation of our model is
that exports and imports are assumed to respond promptly ± within the period
± to changes in their determinants, i.e. the real exchange rate, foreign income
and domestic income. It might be more realistic to assume instead that
imports and exports respond in a distributed-lag fashion.
The third speci®cational change to be considered involves the price
adjustment relation. With the P-bar speci®cation, our model generates a great
deal of persistence in both yt and y~ t , but not much for the in¯ation variable
pt: This absence can be seen in terms of the response of pt to a policy shock
in the lower-left panel in the top half of Figure 1.4; there is some persistence
but not much. The autocorrelation functions for y~ t and pt , which of course
re¯ect responses to all shocks, are as shown in Table 1.3.
From the study of Nelson (1998), it is known that as of 1997 most existing
quantitative models designed to incorporate both sticky prices and optimising
behaviour feature little if any in¯ation persistence.25 One notable exception is
the model of Fuhrer and Moore (1995), which was designed so as to provide a
good match to the US autocorrelation functions for y~ t , pt , and Rt .26
0 0
q
y
–0.5 –5
0 10 20 0 10 20
0 0
–1
s
p
–2 –5
0 10 20 0 10 20
1 1
0 0.5
R
Δp
–1 0
0 10 20 0 10 20
a Responses to unit shock to policy rule, with h = 0.8
0 0
–1
–5
q
y
–2 –10
0 10 20 0 10 20
2 10
1 0
s
p
0 –10
0 10 20 0 10 20
0.4 2
0.2 1
R
Δp
0 0
0 10 20 0 10 20
b Response to unit shock to policy rule, with h = 0.0

a
Figure 1.4 Impulse responses to monetary shock

0.4 0.4
0.1 0.2
q
y
0 0
0 10 20 0 10 20
0 0.2
–0.1 0
p
s
–0.2 –0.2
0 10 20 0 10 20
0 0
Δp
–0.05 –0.02
R
–0.1 –0.04
0 10 20 0 10 20
a Responses to unit shock to IS; h = 0.8
0.2 1
0.1 0.5
y
0 0
0 10 20 0 10 20
0 1
–0.5 0
p
–1 –1
0 10 20 0 10 20
0 0
Δp
–0.5 –0.1
R
–1 –0.2
0 10 20 0 10 20
b Responses to unit shock to IS; h = 0.0
Figure 1.5 Impulse responses to shock to IS

Table 1.3 Autocorrelation functions

Policy parameters: 1 1.0, 2 0, 3 0.8
US dataa Basic model Model with (1.10) replacing (1.2)
Lag y~t pt y~t pt y~t pt
0 1.000 1.000 1.000 1.000 1.000 1.000

1 0.970 .875 .870 .283 .904 .821
2 0.910 .827 .758 .051 .814 .666
3 0.841 .798 .655 ±.013 .726 .531
4 0.769 .776 .567 ±.017 .637 .415
5 0.703 .719 .487 ±.019 .549 .315
Note:
a
Quarterly, 1955:1±1996:4. It should be noted that the the output gap in the ®rst column is
measured as in McCallum and Nelson (1997), not by any detrending procedure based only on the
output series itself.
Consequently, it is of interest to determine how replacement of the P-bar

price-adjustment speci®cation with that of Fuhrer and Moore (F±M) would
affect our model.
In fact, we will adopt a slightly modi®ed version of the F±M speci®cation
with two-period contracts, a version that has been used by Fuhrer (1997) and
Isard, Laxton, and Eliasson (1999), among others. Speci®cally, we consider the
following price adjustment model:
pt 1 � !Et pt1 !pt�1 y~ t ut 1:10
With ! 0.5, this relation is almost identical to the F±M speci®cation, as has
been shown by Walsh (1998, pp. 224±5).27 Here ut re¯ects the random,
unsystematic component of pricing behaviour; it is assumed to be white noise.
For inclusion in our simulation analysis, numerical values must be attached to
and 2u E(u2t ). On the basis of results in Isard, Laxton, and Eliasson (1999),
I have adopted 0.0032 and 2u (0.0025)2 .28
The fourth panel of Table 1.2 reports RMSE values for simulations with the
same policy rule settings as before. As can be seen, the extent to which
in¯ation variability depends upon 1 (our measure of policy activism) is much
less than with the P-bar model. That is because the F±M speci®cation features
much more built-in in¯ation inertia, and also because of the presence in (1.10)
of the ut component that is not present in our P-bar cases. The sensitivity of y~ t
variability to the policy rule is, correspondingly, considerably greater than
with the P-bar model. Autocorrelation coef®cients for pt and y~ t are shown in
the two rightmost columns of Table 1.3. In keeping with our understanding of
the nature of the F±M speci®cation, in¯ation persistence is much greater than
with the P-bar model, and much closer to that found in the US data. The
persistence of y~ t is about the same as in our basic model ± i.e. quite substantial
although less than in the US data. In the ®fth panel of Table 1.2, the policy
feedback rule responds to Et�1 pt1 rather than Et�1 pt , so as to determine
whether this type of `preemptive' response is more effective in the presence of
additional in¯ation inertia. As can be seen, however, the results are not greatly
affected by this change.
Impulse response functions for the F±M pricing speci®cation are shown in
Figures 1.6 and 1.7; the policy rule has 1 1.0 and responds to Et�1 pt1 . The
additional in¯ation inertia provided by this model shows up quite clearly in
the lower left-hand panels. It is worth noting that although our P-bar model
does not give rise to much in¯ation persistence, it does account for a
considerable amount of persistence in output.29 This ®nding con¯icts with
claims made recently by various writers, including Chari, Kehoe and
McGrattan (1995), Christiano, Eichenbaum and Evans (1997) and Andersen
(1998). The reason that such a disagreement is possible is that these authors all
presume that slow adjustment or staggering of goods prices is combined with
continuous clearing of the labour market. But what is assumed in the models
given above ± as well as in the work of McCallum and Nelson (1997,1998),
Taylor (1979, 1993b), Fuhrer and Moore (1995) and many others ± is that ®rms
produce whatever quantity is demanded at the prevailing price with labour
supplying as much labour as is needed (given capital, technology and the
production function). Current wages in this arrangement are irrelevant for
labour quantity determination, except via effects on prices, as in the
ìnstalment payment' discussion of Hall (1980). Labour supply conditions
are important only in the determination of y t , not yt ± y t .30 As a consequence,
these models do not imply that a contractionary monetary shock leads to a rise
in pro®ts, as suggested by Christiano, Eichenbaum and Evans (1997).
One important speci®cational issue that will not be explored quantitatively
concerns the absence of any monetary variables in the basic model of Section
3. In this respect that model is consistent with most recent analysis of
monetary policymaking, as represented in the NBER conference volume edited
by Taylor (1999). But is it actually reasonable to conduct monetary policy
analysis using an analytical framework that includes no money demand
function and indeed no reference to any monetary aggregate, either narrow or
broad? At a super®cial level, this question is answered by the well known point
that if a money demand function were appended to a basic model such as
(1.1)±(1.3), its only role would be to determine how much money would have
to be supplied to implement the interest instrument policy rule; implied paths
of yt , pt and Rt would be entirely unaffected.
At a less super®cial level, however, the question becomes one that asks
whether an optimising speci®cation, with the medium-of-exchange role of
money properly recognised, would yield an expectational IS function that
includes no real money balance terms. The answer to that question is that such
terms are absent only when the implied ìndirect utility function' for the
optimising household is additively separable in consumption and real money
balances. Thus formulations of the expectational IS function of the type that
1 5
0 0
q
y
–1 –5
0 10 20 0 10 20
0 0
–0.5
s
p
–1 –5
0 10 20 0 10 20
0 1
Δp
–0.05 0
R
–0.1 –1
0 10 20 0 10 20
a Responses to unit shock to policy rule
0.2 0
0
q
y
–0.2 –0.05
0 10 20 0 10 20
0.05 0.05
0
p
0 –0.05
0 x10–3 10 20 0 10 20
5 0.01
Δp
0.005
R
0 0
0 10 20 0 10 20
b Responses to unit shock to IS
Figure 1.6 Impulse responses with F±M (1.10)

0.1 2
0 0
q
y
–0.1 –2
0 10 20 0 10 20
0.02 2
0.01 1
s
p
0 0
0 x10–3 10 20 0 x10–3 10 20
5 4
Δp
0 2
R
–5 0
0 10 20 0 10 20
a Responses to unit shock to UIP
1 2
0.5 1
y
0 0
0 10 20 0 10 20
0 2
–0.5 0
s
p
–1 –2
0 10 20 0 10 20
0 0.1
Δp
–0.05 0
R
–0.1 –0.1
0 10 20 0 10 20
bb Responses
Responses toto unit
unit shocktotoytY
shock
–
Figure 1.7 Impulse responses with F±M (1.10)

are prevalent in the literature ± and used above ± depend upon this separability
assumption. To evaluate whether such an assumption is appropriate, one
would have to consider alternative ways of modelling the role of the medium
of exchange ± e.g. money in utility function, shopping time, transaction cost
or cash-in-advance setups ± and alternative functional forms (complete with
quantitative properties). Such a study is far beyond the scope of the present
paper, so I will end this discussion simply by noting that separability is not
compatible with the shopping-time formulation utilised by McCallum and
Goodfriend (1987).
5 Model diagnostics
In the previous section it has been demonstrated that changes in speci®c

details of a structural model can make major differences in its policy-relevant
dynamic properties. Consequently, it is important to have a strategy for
conducting model diagnostics, so as to ascertain readily and reliably which
models or model variants are more nearly consistent with actual macro-
economic data.
In this regard there are clearly many ways to proceed, since there are
alternative ways of presenting the various second moments relevant to a
model's performance. Again I would like to suggest, however, a procedure that
differs from the ones most common in the VAR-oriented literature. More
speci®cally, I would suggest that vector autocorrelation functions of the type
utilised by Fuhrer and Moore (1995) ± but augmented by univariate variance
statistics ± may be a more fruitful source of information than the impulse
response functions that are more frequently emphasised.
The basic point is as follows. There are reasonably robust procedures,
developed by Christiano, Eichenbaum and Evans (1997) and Bernanke and
Mihov (1998), for identifying monetary policy shocks, but these procedures do
not identify the other structural disturbances of a dynamic macroeconomic
model.31 Therefore, these procedures do not automatically provide data
analysis counterparts to be compared with impulse response functions for the
candidate model's structural shocks other than the policy instrument shock.32
By contrast, vector autocorrelation functions for actual data constitute pure
descriptive statistics that can be readily compared with analogous statistics
implied by a candidate model. It is of course true that autocorrelations in two
sets of (actual or hypothetical) data could agree while the autocovariances
nevertheless differed in magnitude. Accordingly, one needs to modify the
Fuhrer±Moore statistics to re¯ect autocovariances rather than autocorrela-
tions. Or, equivalently, one could augment the autocorrelations with
magnitudes of the variances of each variable in the (actual or hypothetical)
data set. To me the latter possibility seems somewhat more attractive, since it
divides the comparison into two parts. The univariate variances indicate
whether the variability of the model's variables matches that in the data, while
Table 1.4 Model and data variability
p y~ R
a US data, variancea 0.36 4.97 0.49

1955:1±1996:4 std dev.b 2.41 2.23 2.80
b Basic model variance 4.11 3.10 0.53

(0.5, 0.4, 0.8) std. dev. 8.11 1.76 2.92
c Basic model variance 2.56 3.13 1.00

But with h 0 std dev. 6.40 1.77 4.00
d Model with variance 0.87 3.52 0.65

(1.10) std dev. 3.74 1.88 3.21
e Model with variance 1.04 2.56 0.94

(1.10) and h 0 std dev. 4.07 1.60 3.89
Notes:
a
Variance statistics are quarterly fractional units multiplied by 104 in all panels.
b
Standard deviations are in percentages, annualised for p and R, in all panels.
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
–
Autocorrelation
Autocorrelationfunctions forΔΔp,yỹ,, RR in
functionsfor in US
US data
date
Figure 1.8 Autocorrelation functions for US data

1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
p, ỹy~,, R
Δp,
Autocorrelation functions for Δ R ininbasic
basicmodel
model
Figure 1.9 Autocorrelation functions for basic model
the autocorrelation magnitudes and patterns re¯ect the nature of the dynamic
interrelationships. Any major discrepancy on any of these dimensions ± any
discrepancy between a model's properties and actual data ± re¯ects a weakness
in the model's speci®cation. This argument presumes, clearly, that the policy
rule in the model simulations and the various shock variances are realistically
matched to the ones that prevailed over the sample period during which the
data was generated.
The foregoing objection to impulse response methods does not pertain, of
course, to VAR systems in which all shocks, not just the one associated with
monetary policy actions, are identi®ed. Examples are provided by Sims
(1998a), Blanchard and Watson (1986) and many others. But the identifying
restrictions in these systems are much more demanding and less credible than
in the semi-structural systems promoted by Bernanke and Mihov (1998a),
Christiano, Eichenbaum and Evans (1998) and others who seek robustness.
But whether this point of view is persuasive or not, the vector autocorrelation
strategy seems at least somewhat attractive because of its purely descriptive
nature (as mentioned by Fuhrer and Moore, 1995). Accordingly, an illustrative
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
Autocorrelation functions for Δp, y~, R in basic model, h = 0.0
Figure 1.10 Autocorrelation functions for basic model (h 0.0)
application will be presented as the remainder of this section.

For the following experiments, we use policy rule (1.9) with parameter
values 1 0.5, 2 0.4 and 3 0.8. These are chosen to be representative of
actual policy behaviour, as estimated by McCallum and Nelson (1998)
following Clarida, Gali and Gertler (1998). We begin by combining this rule
with our basic model and comparing its autocorrelation properties (plus
variances) with those of actual data. For the latter, I use seasonally adjusted
observations over 1955:1±1996:4 on pt , y~ t and Rt as described in McCallum
and Nelson (1998). The three variances ± alternatively reported as annualised
percentage standard deviations ± are shown in panel a of Table 1.4, with
autocorrelations presented in Figure 1.8.
A comparison of panels a and b of Table 1.4 indicates that the basic model
implies variances of a realistic magnitude for y~ and R, but much too large for
p. In addition, the autocorrelations depicted in Figure 1.9 fail badly to match
those of Figure 1.8 in all panels except those for the own autocorrelations of y~
and R. Two of the three contemporaneous correlation coef®cients are of the
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
Autocorrelation functions for Δp, ~

y, R in model with (1.10)
Figure 1.11 Autocorrelation functions for model with (1.10)
same sign in the model and in the data, but only one (for p and R) represents
a reasonably close quantitative match.
The third panel in Table 1.4 and Figure 1.10 pertain to the same model
except with h 0 ± i.e. with habit formation eliminated from the households'
saving decision. Surprisingly, this elimination slightly increases the persis-
tence of in¯ation. But it does not overcome the other major problems with the
basic model.
Next we turn to the model in which the price adjustment (1.10) replaces the
P-bar speci®cation. Now the variance magnitudes, reported in panel d of
Table 1.4, are closer to those in the data. And the own autocorrelation
functions shown in Figure 1.11 are distinctly more similar to those of Figure 1.8.
Indeed, they provide a match that might be judged as semi-respectable. But
the cross autocorrelations match quite poorly, especially those involving y~ .
Thus this chapter's ®ndings are basically consistent with those reported by
Fuhrer (1997, 1998).33 Setting h 0, in Figure 1.12 and panel e of Table 1.4,
worsens the match between model and data, especially in terms of the
variance magnitudes.
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
1 1 1
0 0 0
–1 –1 –1
0 20 40 0 20 40 0 20 40
Autocorrelation functions for Δp, y˜y~,, RR in
Δp, in model
model with
with (1.10) and hh == 0.0
(1.10) and 0.0
Figure 1.12 Autocorrelation functions for model with (1.10) (h 0.0)
What can be concluded from these exercises concerning the usefulness of

the variance plus vector-autocorrelation approach to model diagnostics? The
basic similarity across Figures 1.9±1.12 of several of the panels suggests that the
autocorrelation properties are not as sensitive to a model's speci®cation as are
the impulse response properties. This is admittedly a mark against the former
approach, but arguably one that is not as serious as those against the impulse
response approach that are mentioned on p. 31. The crucial fact, I would
suggest, is that the variances and autocorrelations together are able to indicate
(1) clear discrepancies relative to the actual data and (2) clear differences
among model variants ± both without requiring any (inherently dubious)
identi®cation assumptions other than those used in developing the model.
6 Other approaches
Considerable attention has been devoted, during recent years, to the `narrative
approach' to measuring the effects of monetary policy that was pioneered by
Romer and Romer (RR) (1989). As is well known, the RR study generated a
number of dates at which the Fed is judged to have èxogenously' adopted a

more stringent policy stance in order to reduce in¯ationary pressures. Shapiro
(1994), Leeper (1997) and others have noted, however, that responses to
in¯ation pressures are clearly not exogenous in the sense relevant to the
systematic versus shock decomposition. Shapiro's study represents an
improvement in that regard, but would still seem open to the criticism that
it builds upon a measure of policy actions that differs from the usual ones in
the following three respects: (1) Traditional measures use variables that can
range over a near-continuum of values, rather than only two, so can
distinguish between major and minor actions. (2) The RR dummy-variable
measure recognises as non-zero actions only those in a contractionary direction,
leaving decisions to be unusually expansionary to be included together with
normal behaviour. (3) Values of the RR dummy are based on what it is that the
Federal Open Market Committee's records say, not on what the Open-Market
Desk actually does. All in all, then, it is dif®cult to understand the enthusiastic
reception that this approach has received. In any event, the present chapter is
concerned more with the systematic portion of policy rather than the portion
toward which the RR approach is ostensibly directed.
A striking and unusual analysis was put forward by Sims (1998b). In this
paper, Sims utilises VAR procedures but in a different and bolder fashion than
that ± typi®ed by Bernanke and Mihov (1998) or Christiano, Eichenbaum and
Evans (1998) ± mentioned in Section 2. In particular, Sims (1998b) estimates
an identi®ed VAR system for US monthly data from the interwar period
1919:08±1939:12 and then conducts counterfactual historical simulation
policy analysis by replacing the estimated monetary policy rule ± represented
by the VAR equation explaining movements in the Federal Reserve discount
rate ± with one estimated on postwar data 1948:08±1997:10. The striking
®nding emphasised by Sims is that this replacement has very little effect on
the estimated time path of real output that would have obtained over 1919±
39, given the estimated shocks of that era. In other words, monetary policy as
practised during postwar years would not have prevented the Great
Depression. This dramatic conclusion is reinforced by Christiano's (1999)
®nding that Sims' conclusion is not overturned by replacement of his discount
rate rule with one that generates a much more expansionary time path for the
M1 money stock.
Both of these studies are, clearly, open to Lucas Critique objections. Both
authors recognise that problem, Sims contending that the usual objection is
philosophically ¯awed (1998b, pp. 154±6) and Christiano leaving it for
consideration by his readers (1999, p. 4). My own belief is that the relationship
between macroeconomic variables ± nominal income or prices and output ±
and monetary policy variables during the interwar years cannot be
satisfactorily represented by a model that does not include some variable
representing the effects of ®nancial crises. In my (1990) study, for example, I
found that the relationship between M1 growth and growth in the monetary
base was strongly in¯uenced by a measure of current bank failures (prior to the
creation of the FDIC).34
Another VAR-oriented approach to the study of systematic monetary policy
responses was developed by Bernanke, Gertler and Watson (1997), who
concluded that a large fraction of the US economy's real effects from oil price
shocks since 1970 has resulted from the monetary policy response to these
shocks, rather than from the shocks themselves. The study is concerned with
attributing historical ¯uctuations to various sources, rather than with the type
of characterisation attempted in the present chapter. Bernanke, Gertler and
Watson (1997, p. 93) state that their method `certainly is not invulnerable to
the Lucas critique'.
Closer in spirit and approach to the present chapter is a study by Dotsey
(1999). It, too, utilises simulations with a setup that features maximising
behaviour and aspires to the development of a policy-invariant, structural
model, and it reaches similar conclusions. One major difference relative to the
present chapter is that Dotsey's comparisons are made across entirely different
policy rule speci®cations, rather than across different parameter settings for
variants of a single rule, as is typically the case in Sections 3±5 above.
7 Conclusions
Let us conclude with a very brief summary. The chapter has argued that, in
studying the monetary policy transmission process, more emphasis should
be given to the systematic portion of policy behaviour and correspondingly
less to random shocks ± basically because shocks account for a very small
fraction of policy instrument variability. Analysis of the effects of the
systematic part of policy requires structural modelling, rather than VAR
procedures, because the latter do not give rise to behavioural relationships
that can plausibly be regarded as policy-invariant. By use of an illustrative
structural speci®cation with variants, the chapter characterises the effects of
policy parameter settings by means of impulse response functions and root-
mean-square statistics for target errors. Different models give different
answers to questions about the effects of systematic policy, so procedures for
scrutinising model speci®cation are essential. In this regard, it is argued that
vector autocorrelation functions, augmented by variance statistics for each of
a model's variables, seem more promising than impulse response functions
because the latter require shock identi®cation, which is inherently a dif®cult
process.
Notes
1. I am indebted to Miguel Casares, Larry Christiano, Marvin Goodfriend, Jeffrey
Lacker, Ellen McGrattan, Allan Meltzer, Edward Nelson and Harald Uhlig for
helpful commments and suggestions.
2. Meltzer's contribution to another 1995 symposium entitled `Channels of Monetary

Policy', sponsored by the Federal Reserve Bank of St Louis, focuses instead on the
role of nominal price stickiness.
3. Some examples are Cochrane (1994), Sims (1992), Christiano, Eichenbaum and Evans
(1998), Chari, Kehoe and McGrattan (1996), and Bernanke and Blinder (1992).
4. An extensive and sophisticated surrey of this portion of the literature is provided by
Christiano, Eichenbaum and Evans (1997). Also see Bernanke and Mihov (1998)
and Sims (1992).
5. I thank Richard Clarida for providing me with the relevant standard deviations.
6. It is well known that most models with non-instantaneous price adjustment
behaviour do not satisfy this proposition.
7. The natural rate concept used in this chapter is the value of yt that would prevail if
prices were fully ¯exible ± i.e. if there were no nominal stickiness in the economy.
8. That some fundamental points can be unaffected by this type of neglect is
demonstrated by two examples in McCallum and Nelson (1999). Such is de®nitely
not true in general, however.
9. References are Calvo (1983) and Rotemberg (1982).
10. In particular, the P-bar model satis®es the natural rate hypothesis ± i.e. that E(yt ±
y t ) 0 for any monetary policy rule, which is not the case for the other two
speci®cations.
11. For an extensive discussion of the MSV solution concept, see McCallum (1999).
12. The perverse response of pt to a policy shock is not a general implication of the P-
bar model, as will be seen below.
13. If the central bank is mistaken in its belief about the average value of r, this will
result in an average in¯ation rate that differs from the target .
14. Is there any evidence in approaches oriented toward structural relations such as
(1.1)±(1.3)? Given the system (1.1)±(1.3), there is no hypothesis of this type testable
with data from a single regime. But there is the possibility of testing speci®cations
(1.1) and (1.2) against alternatives that imply the presence of money illusion.
Incidentally, as 1 ! 1 the solution becomes arbitrarily close to one in which Rt
is set so as to make Et�1 pt1 *. This indicates that there is not much difference
between ìnstrument rules' and the `forecast targeting' procedure emphasised by
Svensson (1999).
15. This statement presumes that expectations are rational.
16. Bernanke, Gertler and Watson (1997, p. 92) state that Ìt is not possible to infer the
effects of changes in policy rules from a standard identi®ed VAR system'. Important
references to the literature include Bernanke (1986), Bernanke and Blinder (1992),
Blanchard and Watson (1986), Christiano and Eichenbaum (1992) and Christiano,
Eichenbaum, and Evans (1998).
17. This version is due to Lucas (1972). For a brief discussion, see McCallum and
Nelson (1997).
18. As mentioned above, we treat capital as exogenously determined.
19. The model is calibrated by reference to relationships estimated in various studies
with quarterly data. A value of 0.8 for h was estimated by Fuhrer (1998).
20. I am not entirely happy with this practice, which implicitly attributes knowledge to
policymakers that they could not actually possess.
21. The asymptotic effect on qt is nevertheless the same ± zero ± in the two cases.
22. Although y t depends on raw material inputs, it also has a technology-shock term, as
in the real business cycle literature. This shock process is autoregressive of order 1,
with parameter 0.95.
23. This difference has been usefully emphasised by Fuhrer (1997, 1998). A paper by
Estrella and Fuhrer (1998) stresses that `standard models' with h 0 and a Calvo±
Rotemberg price adjustment speci®cation are seriously inconsistent with the data. I
agree with that judgement but see no reason to conclude that all optimising models
with rational expectations are unsatisfactory in this regard.
24. To close down the basic model, it is necessary not only to eliminate exports and
imports, but also to adjust the variance of the shock term driving y t (because the
latter no longer depends on imported raw materials). For more explanation, see
McCallum and Nelson (1998).
25. To re¯ect price level stickiness, some departure from full optimizing behaviour is
required ± e.g. some additional constraint must be imposed relative to a ¯exible
price general equilibrium system. There is considerable scope for dispute
concerning the relative `rationality' of different departures.
26. The Fuhrer±Moore (1995) paper does not use optimising analysis to generate its
consumption behaviour, but instead posits a non-expectational IS function. Also it
uses detrended yt as its measure of the output gap.
27. The difference is that (1.10) includes only y~ t in place of 0.5Et (y~ t y~ t 1 ).
28. To get 0.0032 from the Isard, Laxton and Eliasson (1999) value of 3.2, one divides
by 400, because they express in¯ation in annualised percentage points, and then
divides by 2.5 to re¯ect the slope of an Okun's Law relationship between y~ t and
unemployment.
29. Simulation results indicate that yt features signi®cantly more persistence than does y~ t .
30. Recall that y t is the natural rate value of yt that would prevail in the absence of any
nominal frictions.
31. There exists some controversy even over the robustness of these procedures. For
recent contributions on this topic, see Faust (1998) and Uhlig (1999).
32. It would be possible to judge a model's ®t entirely on the basis of the impulse
response functions for the policy shock, as in Rotemberg and Woodford (1997), but
this seems undesirable given the small contribution to overall variability coming
from this source.
33. They are also somewhat in the spirit of Estrella and Fuhrer (1998), but do not
involve the Calvo±Rotemberg pricing equation that was a component of the MN
model criticised by Estrella and Fuhrer.
34. My study found that a policy rule that adjusts the monetary base so as to attempt to
keep nominal income on a steady growth path would have made the 1930s' fall in
nominal income much less severe than actually occurred. This activist feedback
rule would have resulted in a much greater expansion of M1 than in Christiano's
counterfactual simulation (which was in turn more expansionary than Sims').
What this monetary stimulus would have done for real output depends, of course,
on the model utilised.
References
Andersen, T. M. (1998) `Persistency in Sticky-price Models', European Economic Review,
42, p. 593±603.
Bernanke, B. S. (1986) Àlternative Explanations of the Money-income Correlation',
Carnegie±Rochester Conference Series on Public Policy, 25, 49±99.
Bernanke, B. S. and A. S. Blinder (1992) `The Federal Funds Rate and the Channels of
Monetary Transmission', American Economic Review, 82, 901±21.
Bernanke, B. S. and M. Gertler (1995) Ìnside the Black Box: The Credit Channel of
Monetary Policy Transmission', Journal of Economic Perspectives, 9, 27±48.
Bernanke, B. S., M. Gertler and M. W. Watson (1997) `Systematic Monetary Policy and
the Effects of Oil Price Shocks', Brookings Papers on Economic Activity, No. 1, 91±142.
Bernanke, B. S. and I. Mihov (1998) `Measuring Monetary Policy', Quarterly Journal of
Economics, 113, 869±902.
Blanchard, O. J. and M. W. Watson (1986) Àre Business Cycles All Alike?', in R. J.
Gordon (ed.), The American Business Cycle, Chicago, University of Chicago Press.
Calvo, G. A. (1983) `Staggered prices in a Utility-maximizing Framework', Journal of
Monetary Economics, 12, 383±98.
Chari, V. V., P. J. Kehoe and E. R. McGrattan (1996) `Sticky Price Models of the Business
Cycle: Can the Contract Multiplier Solve the Persistence Problem', NBER Working
Paper 5809.
Christiano, L. J. (1999) `Discussion', in Beyond Shocks: What Causes Business Cycles?,
Boston, Federal Reserve Bank of Boston.
Christiano, L. J. and M. Eichenbaum (1992) Ìdenti®cation and the Liquidity Effect of a
Monetary Policy Shock', in A. Cukierman, Z. Hercowitz and L. Leiderman (eds) Political
Economy, Growth, and Business Cycles, Cambridge, MA, MIT Press.
Christiano, L. J., M. Eichenbaum and C. L. Evans (1997) `Sticky Price and Limited
Participation Models: A Comparison', European Economic Review, 41, 1201±49.
ÐÐÐÐ (1998) `Monetary Policy Shocks: What Have We Learned and to What End?',
NBER Working Paper 6400.
Clarida, R., J. Gali and M. Gertler (1998) `Monetary Policy Rules in Practice: Some
International Evidence', European Economic Review, 42, 1033±68.
Cochrane, J. H. (1994) `Shocks', Carnegie±Rochester Conference Series on Public Policy, 41,
295±364.
Dotsey, M. (1999) `The Importance of Systematic Monetary Policy for Economic
Activity', Federal Reserve Bank of Richmond Economic Quarterly, 85, 41±59.
Dotsey, M. and P. N. Ireland (1995) `Liquidity Effects and Transaction Technologies',
Journal of Money, Credit, and Banking, 27, 1441±56.
Estrella, A. and J. C. Fuhrer (1998) `Dynamic Inconsistencies: Counterfactual Implica-
tions of a Class of Rational Expectations Model', Federal Reserve Bank of Boston,
Working Paper 98±5.
Faust, J. (1998) `The Robustness of Identi®ed VAR Conclusions about Money', Carnegie±
Rochester Conference Series on Public Policy, 49, 207±44.
Fuhrer, J. C. (1997) `The (UN)importance of Forward-looking Behavior in Price
Speci®cations', Journal of Money, Credit, and Banking, 29, 338±50.
ÐÐÐÐ (1998) Àn Optimization-based Model for Monetary Policy Analysis: Can Habit
Formation Help?', Boston, Federal Reserve Bank of Boston, Working Paper.
Fuhrer, J. C. and G. R. Moore (1995) Ìn¯ation Persistence', Quarterly Journal of Economics,
110, 127±59.
Hall, R. E. (1980) Èmployment ¯uctuations and Wage Rigidity', Brookings Papers on
Economic Activity, 1, 91±123.
Isard, P., D. Laxton and A.-C. Eliasson (1999) `Simple Monetary Policy Rules under
Model Uncertainty', IMF Working Paper 99/75; also in P. Isard, A. Razin and A. K. Rose
(eds) International Finance and Financial Crises, Boston, Kluwer Academic Publishers.
Kerr, W. and R. G. King (1996) `Limits on Interest Rates Rules in the IS Model', Federal
Reserve Bank of Richmond Economic Quarterly, 82, 47±75.
Klein, P. (1997) Ùsing the Generalized Schur Form to Solve a System of Linear
Expectational Difference Equations', University of Stockholm, Working Paper.
Leeper, E. M. (1997) `Narrative and VAR Approaches to Monetary Policy: Common
Identi®cation Problems', Journal of Monetary Economics, 40, 641±58.
Leeper, E. M., C. A. Sims and T. Zha (1996) `What Does Monetary Policy Do?', Brookings
Papers on Economic Activity, 2, 1±63.
Lucas, R. E., Jr. (1972) Èconometric Testing of the Natural Rate Hypothesis', in
O. Eckstein (ed.), Econometrics of Price Determination, Washington, DC, Board of
Governors of the Federal Reserve System.
McCallum, B. T. (1990) `Could a Monetary Base Rule have Prevented the Great
Depression?', Journal of Monetary Economics, 26, 3±26.
ÐÐÐÐ (1999) `Role of the Minimal State Variable Criterion in Rational Expectations
Models', Working Paper, NBER Working Paper 7087; also in P. Isard, A. Razin and A. K.
Rose (eds) International Finance and Financial Crises, Boston, Kluwer Academic
Publishers.
McCallum, B. T. and M. S. Goodfriend (1987) `Demand for Money: Theoretical Studies',
in J. Eatwell, M. Milgate and P. Newman (eds), The New Palgrave: A Dictionary of
Economics, Macmillan, also in J. Eatwell, M. Milgate, P. Newman, The New Palgrave
Dictionary of Money and Finance, London.
McCallum, B. T. and E. Nelson (1997) `Performance of Operational Policy Rules in an
Estimated Semi-classical Structural Model', NBER Working Paper 6599; also in Taylor
(1999).
ÐÐÐÐ (1998) `Nominal Income Targeting in an Open-Economy Optimizing Model',
NBER Working Paper, 6675, also in Journal of Monetary Economics, 43, 553±78.
ÐÐÐÐ (1999) Àn Optimizing IS-LM Speci®cation for Monetary Policy and Business
Cycle Analysis', Journal of Money, Credit, and Banking, 31, 296±316.
Meltzer, A. H. (1995) `Monetary, Credit (and Other) Transmission Processes: A
Monetarist Perspective', Journal of Economic Perspectives, 9, 49±72.
Mishkin, F. S. (1995) `Symposium on the Monetary Transmission Mechanism', Journal of
Economic Perspectives, 9, 3±10.
Nelson, E. (1998) `Sluggish In¯ation and Optimizing Models of the Business Cycle',
Journal of Monetary Economics, 42, 303±22.
Obstfeld, M. and K. Rogoff (1995) `The Mirage of Fixed Exchange Rates', Journal of
Romer, C. and D. Romer (1989) `Does Monetary Policy Matter? A New Test in the Spirit of
Friedman and Schwartz', NBER Macroeconomics Annual 1989, Cambridge, MA, MIT Press.
Rotemberg, J. J. (1982) `Monopolistic Price Adjustment and Aggregate Output', Review of
Economic Studies, 44, 517±31.
Rotemberg, J. J. and M. Woodford (1997) Àn Optimization Based Econometric
Framework for the Evaluation of Monetary Policy', NBER Macroeconomics Annual
1997, Cambridge, MA, MIT Press.
ÐÐÐÐ (1999) Ìnterest Rate Rules in an Estimated Sticky Price Model', NBER Working
Paper 6618.
Rudebusch, G. D. (1998) `Do Measures of Monetary Policy in a VAR Make Sense?'
International Economic Review, 39.
Shapiro, M. D. (1994) `Federal Reserve Policy: Cause and Effect', in N. G. Mankiw (ed.),
Monetary Policy, Chicago, University of Chicago Press.
Sims, C. A. (1992) Ìnterpreting the Macroeconomic Time Series Facts: The Effects of
Monetary Policy', European Economic Review, 36, 975±1000.
ÐÐÐÐ (1998a) `Comment on Glenn Rudebusch's ``Do measures of monetary policy in
a VAR Make Sense?' '', International Economic Review, 39.
ÐÐÐÐ (1998b) `The Role of Interest Rate Policy in the Generation and Propagation of
Business Cycles: What has Changed Since the 30's?', in Federal Reserve Bank of Boston,
Beyond Shocks: What Causes Business Cycles?, Boston, Federal Reserve Bank of Boston.
Svensson, L. E. O. (1999) Ìn¯ation Targeting: Some Extensions', Scandinavian Journal of

Economics, 101, 337±367.
Taylor, J. B. (1979) `Staggered Wage Setting in a Macro Model', American Economic Review,
Papers & Proceedings, 69, 108±13.
ÐÐÐÐ (1993a) `Discretion Versus Policy Rules in Practice', Carnegie±Rochester
Conference Series on Public Policy, 39, 195±214.
ÐÐÐÐ (1993b) Macroeconomic Policy in a World Economy, New York, W. W. Norton.
ÐÐÐÐ (1995) `The Monetary Transmission Mechanism: An Empirical Framework',
Journal of Economic Perspectives, 9, 11±26.
ÐÐÐÐ (1999) (ed.), Monetary Policy Rules, Chicago, University of Chicago Press.
Uhlig, H. (1997) `What are the Effects of Monetary Policy? Results from an Agnostic
Identi®cation Procedure', Tilburg University, Working Paper.
Walsh, C. E. (1998) Monetary Theory and Policy, Cambridge, MA, MIT Press.
Discussion
Lawrence J. Christiano
1 Introduction
Ben McCallum seeks to make two contributions. One is methodological and

the other is more substantive. On the methodological side, the chapter
reviews various empirical strategies that can used for evaluating a monetary
model. On the substantive side, it presents a particular monetary model and
displays a wealth of information about its dynamic properties. Through
numerous well constructed experiments, we learn about the empirical
implications of alternative ways of capturing price rigidity, about the impact
of habit persistence in preferences and about the implications of open
economy considerations. These experiments will be particularly useful for
other researchers working to construct empirically compelling monetary
models.
I will organise my remarks about the methodological theme, though in the
end I will touch on some substantive issues. My comments focus on the
following three topics:
. Monetary policy shocks.
. Second moments and Maximum Likelihood.
. Testing a monetary model using a historical episode.
The perspective that I have on monetary policy shocks differs from that of
McCallum, and I begin my discussion by explaining why. In my discussion of
the use of second moments in empirical analysis, I relate McCallum's analysis
to the existing literature. Finally, in proposing the use of a historical episode to
help evaluate a model, I seek to add another empirical strategy to the set
considered by McCallum. I argue that evaluating a model's ability to account
for a particular episode can be a useful complement to the overall assessment
of a model. The episode I consider is the take-off of in¯ation in the early 1970s,
an experience that I believe represents an important challenge to McCallum's
sticky-price model, and other sticky-price models as well.
44
Discussion 45
2 Monetary policy shocks
To clarify the issues surrounding monetary policy shocks, it is useful to begin

by de®ning a monetary policy rule:
St f
t "t 2:1
Here, St denotes the variable (or, vector of variables) that the central bank
controls directly. McCallum follows much of the recent literature by treating St
as a short-term interest rate. The central bank's decision variable is assumed to
be determined in part as a systematic function, f ; of the state of the economy,
t : But, in practice not all of the variation in St can be accounted for by f

t :
The rest is determined by the monetary policy shock, "t : The literature on
monetary policy shocks seeks to measure "t and to determine its impact on the
major macroeconomic variables. This literature has found that "t accounts for
only a small part of the variation in most variables, including St : McCallum
infers from this that the analysis of monetary policy shocks is therefore not
very useful. He argues that it is much more interesting to study the economic
effects of the systematic part of the monetary policy rule, f
t , and the effects
of alternative speci®cations of St :
But this misses the point of the monetary policy shock literature, at least the
part of that literature with which I am familiar.1 In that literature, the purpose
of measuring monetary policy shocks, and studying their impact on the
economy is precisely to be helpful in the ultimate objective of studying
alternative speci®cations of f and S:
To see why this is so, recall that to study alternative speci®cations of f and S;
one needs a monetary model that can be used as a laboratory for this purpose.
There is no practical alternative. We clearly cannot experiment with
alternative f and Ss on the real world. And, history does not offer observations
on the range of variation in f and S that interest us, for modern economies. So,
to study alternative f and Ss, we need a model. And this is precisely where the
analysis of monetary policy shocks has been helpful. In principle, there are
many models that can serve as laboratories. The analysis of monetary policy
shocks has narrowed the range of models useful for this purpose, and
continues to provide a guide to further development of models.
For example, the class of models called `Lucas misperception' models is
rarely used in quantitative monetary studies today. This is due, in part, to the
®nding of the empirical monetary policy shock literature that the price level
seems to be the last thing ± after employment, output, inventories, etc. ± to
respond to a disturbance in "t : In contrast, the Lucas misperception model
appears to imply that the mechanism by which a disturbance, "t ; impacts the
economy involves as its ®rst step, a change in the aggregate price level.2
Another class of models, naive extensions on the monetary model of Cooley
and Hansen (1989), emphasises anticipated in¯ation effects in the transmis-
46 Lawrence J. Christiano
sion of policy shocks, to the exclusion of liquidity effects. The empirical

monetary shock literature appears to support the view that liquidity effects
play an important role in translating a monetary policy disturbance into
movements in interest rates and money.
The empirical monetary policy shock literature also provides other facts that
are useful as a guide to constructing monetary models. For example, the
®nding that an expansionary monetary policy action does not generate a
signi®cant drop in the real wage severely limits the scope of naive sticky-wage
models. The key mechanism in these models, by which an expansionary
monetary policy shock is translated into an expansion in employment,
operates by reducing the real wage. Similarly, the ®nding that expansionary
monetary policy is associated with a rise in pro®ts limits the scope for naive
sticky-price models, which tend to imply that pro®ts fall with a monetary
expansion.
The analysis of policy shocks has been usefully applied in other areas as well.
For example, Rotemberg and Woodford (1992) analysed the macroeconomic
impact of shocks to government spending to discriminate between alternative
models of government spending. This work has been extended in interesting
ways by Ramey and Shapiro (1998) and Eichenbaum, Fisher and Edelberg
(1999).
Ultimately, we are interested in models which incorporate all the signi®cant
shocks driving the data. Analyses which focus on one shock are particularly
useful as a ®rst step in this enterprise, when the researcher seeks broad
guidance in discriminating between different types of models. As the set of
models worthy of consideration is re®ned, then other methods for model
estimation and diagnosis become relatively more useful. These methods,
which include those preferred by McCallum, involve incorporating all shocks
into a model and comparing a model's implications with the raw data, rather
than just the part driven by one shock or another ± i.e. the impulse response
functions and variance decompositions. This is the phase we are in in the
analysis of monetary models. The set of models in this literature has been
reduced, broadly, to two: models with sticky prices, such as the one studied by
Ben McCallum, and `limited participation' models.3
In sum, there is a limited sense in which I agree with McCallum's opinion
about the role of monetary policy shocks in the construction of models. I am
sympathetic to the idea that it is time to `move on' to analyses which involve
more shocks and which compare a model's implications with the raw data.
However, it is a mistake to minimise the importance of the shock-analysis
literature in this transition. The models we use as we `move on' are models that
were selected from a much larger potential class largely as a result of the
analysis of shocks. Moreover, in the further analysis of models, it is wise for
researchers to bear in mind some of the results from the monetary policy-
shock literature ± for example, the implications of monetary shocks for pro®ts
and real wages mentioned above.
Discussion 47
3 Second moments and Maximum Likelihood
The empirical methods that McCallum prefers involve comparing a model's

implications for the second moment properties of the data with the
corresponding empirical objects. Here, I wish to point out that there is a
time-honoured tradition of using this type of analysis in macroeconomics
generally and in monetary models speci®cally. Moreover, there are further
developments in this type of analysis which could be usefully applied in
analyses like McCallum's.
In a series of papers, Finn Kydland and Edward Prescott have emphasised the
value of evaluating models based on second moment properties. Similarly, a
series of papers by Lars Hansen and Thomas Sargent draws attention to
Gaussian maximum likelihood and other econometric methods based on
second moments in model estimation and testing. In the analysis of monetary
models, Chari, Christiano and Eichenbaum (1995) and King and Watson
(1996) also pursue the strategy of diagnosing models based on an analysis of
second-moment properties of the data.
More recently, there have been interesting new developments in this
literature, which allow a researcher to evaluate a model based on different
frequency components of the data. These include papers by Watson (1993),
Diebold, Ohanian and Berkowitz (1998) and Christiano and Vigfusson (1999).
Often researchers are interested in understanding how different frequency
components of the data impact on model estimation and testing, and this
literature provides the formal tools for doing this. This can be useful for two
reasons. First, understanding which frequencies the model does poorly on can
give guidance to further model development. Second, a researcher may want
to react very differently to a statistical rejection of a model if it arises because of
poor performance in the low frequencies or in the high frequencies. For
example, with a monetary model, one might want to insist particularly that
the model does well in the high frequencies.
4 Testing a model using a historical episode
One way to test a model, not considered by McCallum, is to see how well it
accounts for the events of a particular historical episode. Given the motivation
for the type of work in this chapter, I believe there is one episode that deserves
special attention. As I understand it, a primary motivation for research on
monetary policy rules is to identify strategies for conducting monetary policy
that will ensure that we not repeat the monetary policy mistakes of the past.
One such mistake is the burst of in¯ation experienced by many countries in
the 1970s. In view of this, a natural test of a model like McCallum's is to see
whether it can account for this episode. After all, if the cures for past monetary
disorders emerging from a particular model framework are to be compelling,
that framework must at least be able to provide a plausible account of why

those disorders occurred in the ®rst place.
There is a general class of explanations for the high in¯ation of the 1970s,
according to which it was the result of a weakness either in monetary policy
institutions or in the people in charge of them, that made economies like the
United States vulnerable to a rise in in¯ationary expectations. These
explanations suggest that high in¯ation expectations in the early 1970s ±
triggered perhaps by the commodity and oil shortages of the time ± led central
banks to react with an accommodative monetary policy.4 Consistent with the
observations of Clarida, Gali and Gertler (1998), McCallum's model can
articulate a particular version of this view.
Monetary policy in McCallum's model is represented in the form of (1), in
which St is a short-term rate of interest, Rt ; and
f
t constant 3 Rt�1 1 � 3 1 1 Et t1 2 y t
where t1 is the in¯ation rate in the quarter following quarter t; Et is the
conditional expectation operator and yt is the deviation of output from
potential. A feature of this model is that when 1 < 0; then it is possible for
higher anticipated in¯ation to be self-ful®lling. Since nothing in this result
depends on the value of 3 ; to simplify the intuition I set 3 0: Under these
circumstances, higher expected in¯ation can be self-ful®lling via the following
mechanism.5 Suppose agents expect higher in¯ation. With 1 < 0; this creates
the expectation that the real interest rate will be low, which in turn stimulates
investment demand. The rise in investment demand then triggers a rise in
output and employment which ultimately translates into higher in¯ation,
thereby validating the original jump in in¯ation expectations.
Note that, according to McCallum's model, a burst of high in¯ation
triggered by expectations in this way is associated with a strong economy. A
problem with this argument, however, is that variables like investment and
employment were actually low in the early 1970s, particularly during the 1975
recession.
An alternative model that has been used for monetary analysis is the limited
participation model studied by Christiano (1991), Fuerst (1992) and
Christiano, Eichenbaum and Evans (1998b). Christiano and Gust (1999,
2000) show that this model also has equilibria in which higher expected
in¯ation can be self-ful®lling when 1 < 0: However, Christiano and Gust
(1999b) show that in these equilibria, output and investment are low, a feature
that appears more consistent with the data.
Thus, it appears that the sticky-price model of McCallum fails the test of the
1970s, a test that the limited participation model passes. There is, however,
one important caveat that must be mentioned in this context. There are many
things that happened in the 1970s ± the productivity slowdown began, there
were oil shocks ± and possibly some of the evidence of weakness that
Discussion 49
Christiano and Gust (2000) attribute to self-ful®lling in¯ation expectations is

actually due to these other factors.6 This possibility deserves further
exploration. Still, I suspect that the evidence in fact represents a challenge
to the kind of sticky price model studied by McCallum.
5 Conclusion
In conclusion, McCallum's chapter contains valuable lessons for researchers

constructing quantitative, monetary models for use in analysing the operating
characteristics of monetary policy rules. The focus of McCallum's chapter is on
IS±LM models with sticky prices. McCallum discusses various diagnostic tools
for empirically evaluating these models. Drawing on work currently under-
way with Christopher Gust, I added another tool to the list. The idea is to test
McCallum's model by determining whether it provides a plausible account for
the burst of in¯ation observed in the early 1970s. There is some reason to
think that it cannot. The account articulated by the model for this episode
appears inconsistent with the weakness in employment observed at the time,
particularly during the 1975 recession. This test suggests that an alternative
framework for the analysis of monetary policy rules, the one based on the
limited participation model of money, may be superior.
Notes
1. This section summarises arguments in Christiano, Eichenbaum and Evans (1998a).
2. For a discussion of this model, see Cooley and Hansen (1997).
3. See Fuerst (1992); Christiano (1991); Christiano, Eichenbaum and Evans (1998) and
Alvarez, Atkeson and Kehoe (1999).
4. For an informal discussion of how this might have happened, see Blinder (1982),
especially p. 264. Formal analyses appear in Chari, Christiano and Eichenbaum
(1998) and Clarida, Gali and Gertler (1998).
5. The observation that 1 < 0 can result in equilibria in which in¯ation expectations
are self-ful®lling in a simpli®ed version of McCallum's model (an IS±LM model with
rational expectations) was ®rst made in Kerr and King (1996). Ben McCallum was
kind enough, in a private communication, to con®rm that this result also holds in
the models analysed in his chapter.
6. In Christiano and Gust (2000)'s analysis, they simulate Clarida, Gali and Gertler's
(1998) model and a limited participation model under the assumption that the jump
in in¯ation expectations in the early 1970s was a reaction to a bad technology
shock. This is consistent with the view that the higher in¯ation expectations at the
time were in part due to the rise in oil and other commodity prices that occured in
the wake of the shortages that occurred at the time. The policy rule used in the
simulations uses the parameter values estimated by Clarida, Gali and Gertler (1998)
using data from the 1960s and 1970s.
References
Alvarez, F., A. Atkeson and P. Kehoe (1999) `Money and Interest Rates with
Endogenously Segmented Markets', NBER Working Paper, 7060.
Blinder, A. (1982) Ànatomy of Double-digit In¯ation in the 1970s', in R. Hall (ed.),

In¯ation: Causes and Effects, National Bureau of Economic Research, Chicago,
University of Chicago Press.
Chari, V.V., L. J. Christiano and Martin Eichenbaum (1995) Ìnside Money, Outside
Money, and Short-term Interest Rates', Journal of Money, Credit, and Banking, 27,
November, 1354±1401.
ÐÐÐÐ (1998) Èxpectation Traps and Discretion', Journal of Economic Theory, 81, 2,
462±92.
Christiano, L. J. (1991) `Modeling the Liquidity Effect of a Money Shock', Federal Reserve
Bank of Minneapolis' Quarterly Review, 15, 3±34.
Christiano, L. J., M. Eichenbaum and C. Evans, (1998a) `Monetary Policy Shocks: What
Have We Learned and to What End?', NBER Working Paper 6400; also in J. B. Taylor and
M. Woodford (eds), Handbook of Macroeconomics, Amsterdam, Elsevier.
ÐÐÐÐ (1998b) `Modeling Money', NBER Working Paper, number 6371.
Christiano, L. J. and C. Gust (1999), `Taylor Rules in a Limited Participation Model',
NBER Working Paper 7017.
ÐÐÐÐ (2000) `The Expectations Trap Hypothesis', Federal Reserve Bank of Chicago
Economic Perspectives, 25, 2, 21±39.
Christiano, L. J. and R. J. Vigfusson, (1999), `Maximum Likelihood in the Frequency
Domain: A Time to Build Example', NBER Technical Working Paper 7027.
Clarida, R., J. Gali and M. Gertler (1998) `Monetary Policy Rules and Macroeconomic
Stability: Evidence and Some Theory', New York University, manuscript.
Cooley, T. F. and G. D. Hansen (1989) `The In¯ation Tax in a Real Business Cycle Model',
American Economic Review, 79, 733±48.
ÐÐÐÐ (1997) Ùnanticipated Money Growth and the Business Cycle Reconsidered',
Journal of Money, Credit and Banking, 29, 624±48.
Diebold, F. X., L. E. Ohanian and J. Berkowitz (1998) `Dynamic Equilibrium Economies:
A Framework for Comparing Models and Data', Review of Economic Studies, 65,
433±51.
Eichenbaum, M. J. Fisher and W. Edelberg (1999) Ùnderstanding the Effects of a Shock
to Government Purchases', Review of Economic Dynamics, 2, 166±206.
Fuerst, T. (1992) `Liquidity, Loanable Funds, and Real Activity', Journal of Monetary
Economics, 29, 3±24.
Kerr, W. and R. King (1996) `Limits on Interest Rate Rules in the IS±LM Model', Federal
Reserve Bank of Richmond Economic Quarterly, 82, 47±75.
King, R. G. and M. W. Watson (1996) `Money, Prices, Interest Rates and the Business
Cycle', Review of Economics and Statistics, 78, 35±53.
Ramey, V. and M. Shapiro, (1998) `Costly Capital Reallocation and the Effects of
Government Spending', Carnegie-Rochester Conference Series on Public Policy, 48(1),
45±94.
Rotemberg, J., and M. Woodford (1992) Òligopolistic Pricing and the Effects of Aggregate
Demand on Economic Activity', Journal of Political Economy, 100, 1153±1297.
Watson, M. (1993) `Measures of Fit for Calibrated Models', Journal of Political Economy,
101, 1011±41.
Discussion
Harald Uhlig
1 Introduction
How should monetary policy be conducted? Academics and central bankers

often seem to differ in their opinions. Academics typically advocate openness
(e.g. the publishing of minutes), whereas central bankers often emphasise the
need for secrecy. Academics want central banks to be predictable, whereas
central bankers like to keep the markets on their toes. Central bankers prefer
what Alan Blinder (1989) called ènlightened discretion', whereas academics
often advocate the use of rules. The word `rule' here is to be understood
broadly, and simply means that ®nancial markets can ®gure out what and how
much the central bank is going to do when and under which circumstances.
Put differently, a rule is a systematic framework for predictability. The concept
thus encompasses, for example, the in¯ation-targeting approach in the form
advocated by Lars Svensson (Chapter 2 in this volume). A rule also allows
central bankers to precommit and therefore not to yield to the temptation of
doing something with short-run gains and long-run pains (e.g. lower
unemployment now at the price of higher in¯ation later): a point due to the
famous paper by Kydland and Prescott (1977) and apparently now deeply
ingrained in the hearts, minds and actions of practical central bankers, but
nonetheless worth stating again from time to time.
The continued exchange of academics and central bankers is necessary.
Academics need to understand better why central bankers dislike some of their
suggestions. For example, are there good political or economic reasons for
secrecy? Is the analysis of secrecy provided by academics so far (see Cukierman
and Meltzer, 1986, for example) already complete? Conversely, central bankers
need to understand academics better, too, if for no other reason than to
understand more deeply what it is they are doing and for conducting even wiser,
more focused and further forward-looking policy as a result. Yes, there are good
reasons for following rules above and beyond their use as a commitment device.
Often, `simple' rules such as money growth rules or the Taylor rule have
been advocated, be it for enhancing the communication with the broader
51
52 Harald Uhlig
public, for framing the debate or for providing a ®rst, `rough' benchmark of
comparison. Opponents like to use them to caricature the academic point of
view. But these rules are really meant as a starting point for the analysis and its
debate. Chapter 1 by Ben McCallum is a insightful and pathbreaking
contribution to that end.
The chapter provides an outline of a methodology for the analysis of
monetary transmission mechanisms. Together with a few other leading
researchers, McCallum aims at applying modern advances in macroeconomic
theory to practical issues of monetary policy. One can brie¯y summarise that
methodology as follows. First, write down one or better several models of the
economy with solid microfoundations. Second, show with a careful
comparison with the data, that the model indeed captures the observed
behaviour of the economy. Third, experiment with suggestions for policy
changes, using the models at hand. Fourth, provide a detailed, insightful and
understandable account of the effects of policy and their changes to the policy
maker. Ultimately, the aim must be to provide the policy maker with a deeper
intuition about the cause and the effects of monetary policy and their changes.
Rather than discuss the methodology in the abstract, McCallum provides a
speci®c application, studying the Taylor rule in the context of a model
framework developed by him and Nelson in a number of previous
contributions.
I have little doubt that McCallum's pioneering and ambitious contribution
points us in the right direction. I liked the chapter a lot. If there are any
disagreements, then they are with the details of the arguments and with the
quality of the theoretical advances achieved so far. As a discussant, my role is
to emphasise these points of departure rather than to restate the compara-
tively larger parts where author and discussant agree. In summary: the
approach is a complement not a substitute to other approaches, in particular
the VAR-based methodology, and the theory needs further development.
2 VARs versus theory
A simple view of monetary policy is to think of setting interest rates Rt as a

(systematic) reaction to circumstances plus an unpredictable component,
Rt f past and available present data t
Vector-autoregressions (VARs) identify the monetary policy shock t in the

data and study its effects. This is useful, as they inform us about the effects of
policy surprises on the economy and thus about the transmission mechanism
in general. That literature has made great advances and provided deep insights
in recent years. For example, there is now large agreement that these shocks
are far from being a major source of business cycle ¯uctuations. Obviously,
they shouldn't be if monetary policy is indeed successful in being predictable.
Discussion 53
Theories with solid microfoundations, on the other hand, are more suitable
to contemplate changes in the systematic component f . One may wonder,
whether the systematic component has any effect at all other than in¯uencing
the rate of in¯ation. One may also wonder whether there truly are changes in
the systematic component: perhaps, the continuous trial-and-error evolution
of the way monetary policy responds to current circumstances should more
appropriately be understood to be part of the shock component t . But be this
as it may: one certainly has to conclude that VARs and economic theory are
complements, not competitors. Different problems require different tools.
Good, advanced theories of money are still remarkably hard to come by. There
is a simple reason: this is a very tough objective! Further advances here are not
only desirable, they are crucial.
3 The Model
The results in the chapter are based on an analysis of the model in McCallum
and Nelson (1998). Despite the contrary appearance in chapter 1, the model is
constructed according to the current standards in modern macroeconomic
theory ± i.e. is of the stochastic, dynamic, quantitative, rational expectations
applied dynamic general equilibrium variety. I, for one, ®nd it unfortunate,
that the deep beauty and insight of these models apparently still has to be
`sold' using the now outdated and cruder IS±LM language, but perhaps this is
still necessary for another decade or two to reach the broadest audience
possible. McCallum has rightly shown and emphasised that the IS±LM logic is
not in con¯ict with these models, and, indeed, it isn't. The inherent logic
provides an improvement on the IS±LM view, however. To take a perhaps
strained analogy: there also is no con¯ict between the view that most animals
are near-perfect at the tasks they do as if created by divine intervention for that
purpose, and the view that this near-perfection is the result of genetic,
Darwinian evolution: the latter view simply is more helpful for understanding
what we see and for predicting what will happen if the environment changes.
Likewise, a well built model in the modern tradition is going to be reliable and
more helpful for analysing policy changes than the outdated and inferior IS±
LM framework.
The model by McCallum has a number of special features. The economy is
modelled as open. Investment is exogenous. Labour supply and thus output
gaps are demand-driven. Firms may ®nd themselves producing below
capacity. There is habit formation in consumption.
4 Exogenous investment?
To assume investment to be exogenous for the purpose of studying the effects

of monetary policy is, of course, somewhat odd. Investment (including
durable consumption) is the most volatile component of output and arguably
54 Harald Uhlig
the most interesting one for monetary policy. Whether directly through
interest rates or indirectly through the credit channel, monetary policy mainly
affects investment.
The effect on consumption via a `price channel' is much too sluggish to be of
major importance. Of course, having a more explicit role for investment can
be done in the context of this model. It makes the model less elegant, but not
much more complicated, but it would then have an additional and appealing
channel for monetary policy. A key dif®culty here would be to reconcile the
observed ¯uctuations in interest rates with the observed ¯uctuations in
investment, a point not suf®ciently emphasised by the current real business
cycle literature.
5 Theory versus Data
Jeff Fuhrer (1997) writes:
It is too easy to write an elegant theoretical model, and too dif®cult to write
a model that also replicates key dynamic elements found in the data.
Policymakers will rightly be leery of the former, and at least somewhat more
comfortable with the latter.
It would be hard to disagree with this statement. How well does McCallum's
model do? McCallum's Chapter 1 invites the reader to investigate this issue. It
provides a number of comparisons to the data, that are similar in spirit to the
practice in the real business cycle literature of comparing (®ltered) correlations
at leads and lags, although more could be provided: Figure 1.8 (p. 32) offers
only a partial assessment. There is also no principal difference between
comparing raw correlation patterns versus comparing impulse response
functions in (reduced-form) VARs: both VARs and correlation patterns provide
summary information about the stochastic properties of the models and the
data-and, in fact, one can essentially compute the former from the latter and
vice versa. Which representation one ®nds more useful thus seems to be purely
a question of taste, and I would have found it helpful to also provide a
systematic comparison of the latter.
The comparison to the data in Tables 1.3, 1.4 and Figures 1.8±1.12 (pp. 27±
36) shows that the differences to the data are still uncomfortably large. Policy
makers should still be leery of the model at hand. That said, one should
understand how dif®cult an exercise this is. While we are not at the end of the
road yet, the chapter nonetheless provides a big step forward. Much to his
credit, McCallum discusses all these issues openly and frankly rather than
hiding them for someone else to ®nd. This is an honest, scienti®c approach
which should serve as the model for how things should be done.
Table 1.4 shows particularly big differences with the data regarding the
in¯ation variability, except for model d. Should the reader conclude that one
Discussion 55
should prefer the Fuhrer±Moore framework over P-bar? Comparing Figures

1.9±1.12 shows that the second-order properties of the models do not seem to
differ very much. How broad and robust really is the range of theoretical
insights gained here?
Some of the other ®gures show some puzzling light on McCallums model.
Figure 1.1b (p. 20) shows prices to go down rather than up after a taste shock
makes consumers want to consume more today. Comparing Figure 1.3a with
Figure 1.3b (p. 22) shows that the path of interest rates does not matter for the
path of output: output follows the same path, whether interest rates go up or
go down. More intuition for why all this is happening, and whether or not this
is reasonable, would be desirable.
Some potential, additional problems may be present, but can be judged less
clearly from the tables and ®gures shown. For example, demand-driven labour
movements are known to easily produce counterfactual countercyclical
productivity movements, an old problem of Keynesian-style macroeconomic
theories. In McCallum's model, there are productivity shocks, too. But still:
what is the net effect?
6 Is there a `price puzzle'?
The role of habit formation in McCallum's model is an interesting one.

The behaviour of the model without habit formation can be glanced from
Figure 1.4b: it produces a `price puzzle' in that prices rise after a contractionary
monetary policy shock. Introduce habit formation, and it disappears (see
Figure 1.4a). While habit formation may be a very reasonable feature of
preferences, it is somewhat ironic to see both the empirical VAR literature as
well as the theory here struggle with making the `price puzzle' disappear.
Should one perhaps contemplate the alternative that indeed prices truly rise
rather than fall after a contractionary shock? There is a simple argument for
possibly taking that alternative seriously at least for a small open economy.
The argument has been put forth by Fabio Ghironi (1999) and runs as follows.
Let R; R denote domestic and foreign interest rates. Let e denote the change
in exchange rates, with an appreciation of the domestic currency indicated by
e > 0. Let ; denote domestic and foreign in¯ation. If purchasing power
parity (PPP) holds, then
e �
If uncovered interest parity (UIP) holds, then
R R � e
Taken together, one ®nds
R � R
56 Harald Uhlig
indicating that there then has to be a `perverse' reaction of in¯ation to

monetary policy: holding and R constant, in¯ation will rise one-for-one
with a rise in the domestic interest rate R, i.e. in a monetary contraction! Of
course, it is well known, that neither PPP nor UIP holds well in the data but
still: these are key benchmark concepts in the theory of international trade.
And the view that in¯ation slows down with a rise in interest rates is a key
pillar of our understanding of monetary policy. The simple argument above
shows that at least one of the three has to yield. More broadly, our
understanding of the liquidity effect of monetary policy is still more patchy
than our understanding of Fisherian relations between real and nominal
interest rates.
7 Comparing alternative policy rules
For monetary policy and its effect on the economy, is there a sizeable
difference between reacting to current in¯ation Et�1 pt versus reacting to
in¯ation forecasts, Et�1 pt1 ? It is here where the chapter makes a direct
contribution to an intense and current policy debate of direct practical
consequence: the answer sheds important light on the question, whether
monetary policy should be targeting in¯ation forecasts or whether it should
react to current developments in prices.
Table 1.1 seems to provide the answer. Indeed, McCallum writes, that the
stabilising effect of monetary policy on the in¯ation rate is greater when
Et�1 pt , rather than Et�1 pt1 , is the variable responded to. Obviously, this
is true in the sense of tracing out the results of changing the numerical value in
the policy rule. But is it true in the relevant economic sense? Figures D1.1 and
3.5
3
2.5
E(t–1
t –(Δ1p) t)Infl(
ast) as
Output variability
2 target
1.5 t –(Δ1p) t+1
E(t–1 ) t + 1)
Infl(
1 as target
0.5
0
0 5 10 15
Inflation variability
Figure D1.1 Comparison of two rules: the `Taylor menu': in¯ation and output
Figure D1.1 shows the `Taylor menu', juxtaposing the variance of in¯ation with the
variance in output, using the numbers from the ®rst two rows in Table 1.1 in the
chapter. As one can see, the results form a common line.
Discussion 57
4
Interest rate variability
t –(Δ1p) t)Infl(
E(t–1 ast) as
3 target
2 t –(Δ1p) t+1
E(t–1 ) t + 1)
Infl(
as target
0 5 10 15
Inflation variability
Figure D1.2 Comparison of two rules: in¯ation versus interest rates

Figure D1.2 juxtaposes the variance of in¯ation and the variance in interest rates, using
the numbers from the ®rst two rows in Table 1.1 in the chapter. As in Figure D1.1, they
form a common line.
D1.2 provide the answer, using the numbers provided by McCallum in

Table 1.1. Figure D1.1 shows the implied tradeoff between the variability
between in¯ation and output in McCallum's models. I would like to call this
®gure the Taylor menu, since focusing on this tradeoff has been popularised by
John Taylor. In any case, Figure D1.1 shows that the numbers of the ®rst two
rows (I didn't try all three) in Table 1.1 line up beautifully on a line: the only
difference between reacting to Et�1 pt rather than Et�1 pt1 lies in the
speci®c numerical value for 1 one has to choose in order to pick a particular
point on this Taylor menu, but this is a question of technical execution of the
policy rather than any particularly interesting difference (and can easily be
understood as being due to the endogeneity of expectations when focussing
on in¯ation forecasts). Figure D1.2 shows the same result when focusing on
the tradeoff between in¯ation variability and interest rate variability.
I ®nd this similarity to be a more striking and interesting result than the
differences. Perhaps, the entire debate about whether monetary policy should
target in¯ation forecasts or whether it should react to current price
developments ± or, more broadly, perhaps the entire debate about the
particular form of the policy rule to choose ± is largely a vain one. The
discipline of examining the economic consequences is the relevant perspec-
tive. McCallum's chapter does this: indeed a theoretical framework like his is
necessary to sort out `religious' differences from practical consequences. I read
his results as indicating, that the practical differences are small. This might be
an extremely important and interesting result, if it holds up robustly across
58 Harald Uhlig
many models. Central bank researchers should therefore devote substantial

resources to investigating whether it does.
Finally, McCallum's framework offers an opportunity which only models
with complete micro foundations can offer: since preferences of the
consumers are made explicit, one can evaluate the welfare consequences of
different policy choices and one can search for the optimal policy rule and the
optimal policy parameters within any given class. It is unfortunate that the
chapter is silent on this issue, but this will surely become a standard aspect of
this literature as it develops further.
8 The future
Central banks need to analyse the effects of policy changes. Insights into the
effects of changes to the systematic `feedback' part of policy are particularly
important: how should monetary policy react to developments in the
economy? Providing substantive answers requires carefully spelled out
theories, in which potential policy changes can be examined. These theories
should make use of the tools of modern macroeconomics. They should be
stochastic, dynamic, quantitative, rational expectations applied general
equilibrium models. McCallum is one of the leading pioneers of that new
direction, paving the way. These theories are advanced enough that they may
soon be a useful guide for policy making. McCallum's chapter is a big step in
the right direction. Analyses of this type should become the standard, by
which monetary policy is conducted and assessed in the future. And the
discussion should become more informed, moving away from the more
primitive IS±LM language to the more elegant language of applied general
equilibrium analysis.
Indeed, this is already happening in the United States. To an outside
observer, it is remarkable how far and how fast the various branches of the
Federal Reserve System in the United States have moved in building up their
expertise for analysing monetary and economic phenomena. By providing
suf®cient freedom and resources to actively participate in frontier research on
economics and monetary policy, and by encouraging publication in leading
journals and presentations of ®ndings at learned meetings, the research
departments at the board of the Federal Reserve System as well as at many
regional Federal Reserve Banks have provided an environment so attractive
that it has enabled them to recruit the best talent available. They now rival the
leading economics departments at US universities in quality. There is a
friendly, healthy competition between the various branches within the
Federal Reserve System, keeping the analysts at each place honest and working
hard to provide the best analysis possible, with the largest research department
at the centre of the system. The relevance for monetary policy decision making
is obvious: the weight and the clout of the members of the Board of Governors
of the Federal Reserve System increases directly with the weight and the clout
Discussion 59
of the research department to which they have access, with Alan Greenspan's
weight particularly large. For the public, the result has been the best monetary
policy that the United states has ever enjoyed. Obviously, one cannot ascribe
the entire success and the in¯uence of individual board members to the
strength of the research departments. But one shouldn't run the risk of
underestimating it either.
It is hard to imagine that a similar logic will not hold true for the European
System of Central Banks as well. Indeed, it rather seems that the race to the top
has already begun. The economists at these central banks will use the best
available tools to study the important monetary policy questions at hand.
They will therefore be students of McCallum's work for a long time to come.
References
Blinder, A. S. (1989) Central Banking in Theory and Practice, Cambridge, MA, MIT Press.
Cukierman, A. and A. H. Meltzer (1986) À Theory of Ambiguity, Credibility, and
In¯ation under Discretion and Asymmetric Information', Econometrica, 54, 1099±
1128.
Fuhrer, J. (1997) Àn Optimization-based Econometric Framework for the Evaluation of
Monetary Policy ± Comment', NBER Macroeconomics Annual 1997, 346±55.
Ghironi, F. (1999) Àlternative Monetary Rules for a Small Open Economy: The Case of
Canada', University of Berkeley, Department of Economics, November, draft.
Kydland, F. E. and E. C. Prescott (1977) `Rules Rather than Discretion: The Inconsistency
of Optimal Plans', Journal of Political Economy, 85, 473±92.
McCallum, B. T. and E. Nelson (1998) `Nominal Income Targeting in an Open-economy
Optimizing Model', NBER Working Paper, 6675; also in Journal of Monetary Economics,
43.
2
Price Stability as a Target for Monetary

Policy: De®ning and Maintaining Price
Stability1
Lars E. O. Svensson
1 Introduction
The purpose of this chapter is to provide an up-to-date discussion of monetary

policy with `price stability' as the primary objective. The chapter discusses how
`price stability' can be de®ned, and how price stability can be maintained in
practice. It also discusses some lessons for the Eurosystem.
De®ning price stability involves deciding between price-level stability and
low (including zero) in¯ation, choosing the appropriate price index and
selecting the appropriate level for a quantitative target. It also involves
deciding on the role of real variables, like output, in the objectives for
monetary policy. Thus, de®ning price stability boils down to de®ning the
monetary policy loss function.
Maintaining price stability involves meeting the objectives of price stability
± that is, minimising the monetary policy loss function. I consider three main
alternatives, namely commitment to a simple instrument rule (for instance, a
commitment to following a Taylor rule), forecast targeting (for instance,
in¯ation forecast targeting) and intermediate targeting (for instance, monetary
targeting). A sizeable part of the literature on monetary policy seems to focus
on the properties of optimal and simple reaction functions for monetary
policy, like the performance of Taylor-type reaction functions (that is, linear
reaction functions responding to deviations of in¯ation from an in¯ation
target and to the output gap). This literature provides considerable insights
into the characteristics of optimal monetary policy and the properties of
different reaction functions, and thereby provides considerable guidance and
benchmarks for actual monetary policy, but I argue that a commitment to any
of these reaction functions is, for several reasons, neither a good nor a practical
way of conducting monetary policy. Such commitment is not a substitute for a
systematic operational framework for policy decisions by central banks.
Instead, I believe forecast targeting provides such a systematic and
operational framework. Indeed, I believe that the current best practice of
conducting real world monetary policy can be interpreted as the application of
60
De®ning and Maintaining Price Stability 61
forecast targeting. Thus, most of this chapter is a discussion of forecast

targeting. I examine its theoretical background and how, in practice, it can
incorporate judgemental adjustments and extra-model information, the role
of different indicators (including indicators of `risks to price stability') and, in
particular, how it can incorporate complications like non-linearity and model
uncertainty.
The discussion of forecast targeting builds on Svensson (1999b). The new
elements include a more explicit discussion of policy multipliers, judgemental
adjustments, the choice between mean, median and mode forecasts, and the
role of indicators (the latter builds on Svensson and Woodford, 2000). In
particular, I discuss forecast targeting under non-linearities, nonadditive
uncertainty and model uncertainty, and the related generalisation of what can
be called mean forecast targeting to distribution forecast targeting.
Intermediate targeting, in particular monetary targeting, is treated fairly
brie¯y, for several reasons. The recent interest in monetary targeting has
mainly been motivated by the view that monetary targeting is the reason
behind the Bundesbank's outstanding record on in¯ation control and the
possibility that the Eurosystem would choose monetary targeting as its
monetary policy strategy. However, with regard to whether monetary
targeting lies behind Bundesbank's success, ± as discussed, for instance, in
Svensson (1999d) a number of studies of the Bundesbank's monetary policy,
by both German and non-German scholars, has come to the unanimous
conclusion that, in the frequent con¯icts between stabilising in¯ation around
the in¯ation target and stabilising money growth around the money growth
target, the Bundesbank has consistently given priority to the in¯ation target
and disregarded the monetary target.2 Thus, the Bundesbank has actually been
an in¯ation targeter in deeds and a monetary targeter in words only.
Furthermore, although the Eurosystem has adopted a money-growth indicator,
it has strongly rejected monetary targeting as a suitable strategy, on the grounds
that the relation between prices and money may not be suf®ciently stable and
that the monetary aggregates with the best stability properties may not be
suf®ciently controllable (see Issing, 1998). (Furthermore, an extensive and
convincing discussion some twenty-®ve years ago concluded that intermedi-
ate targeting was generally inferior ± see, for instance, Kareken, Muench and
Wallace 1973, Friedman, 1975 and Bryant, 1980.)
The discussion of the lessons for the Eurosystem builds on Svensson (1999d)
The new elements includes further scrutiny of Eurosystem arguments in
favour of its money growth indicator and against in¯ation-forecast targeting.
In discussing monetary policy strategy, as in Svensson (1999d), I ®nd it
helpful to distinguish two of its elements, namely the framework for policy
decisions and communication. By the `framework for policy decisions', I mean
the monetary policy procedures inside the central bank, which, from
observations of various indicators, eventually result in decisions about the
central bank's instruments ± in short, the principles for setting the
62 Lars E. O. Svensson
instruments (which, in the Eurosystem's case, will be a two-week repurchase

rate). By `communication', I mean the central bank's way of communicating
with outsiders (the general public, the ®nancial market, governments, policy-
makers and policy-making institutions which, in the Eurosystem's case,
includes EU institutions and national governments and parliaments).
Communication is part of the implementation of monetary policy, in that it
affects the ef®ciency of monetary policy by, for instance, in¯uencing
expectations, predictability and credibility. Communication also in¯uences
how transparent policy is, which is crucial for central bank incentives and for
accountability and arguably also for the political legitimacy of the policy.
In terms of the distinction between decision framework and communica-
tion, this chapter almost exclusively deals with the decision framework. I have
extensively discussed communication and transparency in in¯ation targeting
in Svensson (1997a) and (1999b) and the same issue with regard to the
Eurosystem in Svensson (1999d). The concluding Section 5 includes some
comments on transparency and forecast targeting.
A large part of the monetary policy literature uses the concept of `rules' in
the narrow sense of a prescribed reaction function for monetary policy. As in
previous papers, for instance Svensson (1997a) and (1999b), I ®nd it helpful to
use monetary policy rules in a wider sense, namely as à prescribed guide for
monetary policy'. This allows ìnstrument rules', (prescribed reaction func-
tions), as well as `targeting rules' (prescribed loss functions or prescribed
conditions that the target variables, or forecasts of the target variables, shall
ful®l).
Furthermore (as in Cecchetti, (1997), for instance) `targeting' here refers to
loss functions and `target variables' to variables in the loss function. Thus
`targeting variable Yt ' means minimising a loss function that is increasing in
the deviation between the variable and a target level. In contrast, in some of
the literature `targeting variable Yt ' refers to a reaction function where the
instrument responds to the same deviation. As discussed in Svensson (1999b
section 2.4) and (1999a), these two meanings of `targeting variable Yt ' are not
equivalent. `Responding to variable Yt ' seems to be a more appropriate
description of the latter situation.
Section 2 discusses the de®nition of price stability, Section 3 discusses
maintaining price stability, Section 4 discusses lessons for the Eurosystem and
Section 5 presents some conclusions.
2 De®ning price stability
Price-level stability versus low in¯ation

How to de®ne `price stability'? The most obvious meaning of `price stability'
would seem to be a stable price level, `price-level stability'. Nevertheless, in
most current discussions and formulations of monetary policy, price stability
instead means a situation with low and stable in¯ation, `low in¯ation'
(including zero in¯ation). The former de®nition implies that the price level is
stationary (or at least trend-stationary). The latter de®nition implies base drift
in the price level, so that the price level will include a unit root and be non-
(trend-)stationary. Indeed, the price-level variance increases without bound
with the forecast horizon. Thus, to refer to low in¯ation as price stability is
indeed something of a misnomer.
Let me refer to a monetary policy regime as price-level targeting or in¯ation
targeting, depending upon whether the goal is a stable price level or a low and
stable in¯ation rate. We can represent (strict)3 price-level targeting with an
intertemporal loss function
X
1
Et Lt; 2:1
0
to be minimised, where (0 < < 1) is a discount factor and the period loss
function is the quadratic loss function
1
Lt pt � pt 2 2:2
2
Here, pt denotes the (log) price level in period t and pt denotes the (log) price-
level target. The price-level target could be a constant or a (slowly) increasing
path,
pt pt�1 2:3
where 0 is a constant (low or zero) in¯ation rate.4 Similarly, we can

represent (strict) in¯ation targeting with a period loss function given by
1
Lt t � 2 2:4
2
where t pt � pt�1 denotes (the) in¯ation (rate) and denotes a low (or
zero) in¯ation target.
Following Cecchetti (1997), we can use more compact notation by
representing in¯ation targeting in (2.2) by the state-dependent price-level
target
pt pt�1 2:5
instead of (2.3.), or by representing price-level targeting in (2.4.) by the state-

dependent in¯ation target
t pt � pt�1 2:6
instead of a constant . Hence, (2.5.) illustrates the base drift in in¯ation

targeting; the in¯ation target applies from the realised price level pt�1 rather
than from the target price level pt�1 . Similarly, (2.6) illustrates that the
in¯ation target becomes endogenous and time-varying under price-level
targeting.
In the real world, there are currently an increasing number of monetary
policy regimes with explicit or implicit in¯ation targeting, but there are no
regimes with explicit or implicit price-level targeting. Whereas the Gold
Standard may be interpreted as implying implicit price-level targeting, so far
the only regime in history with explicit price-level targeting occurred in
Sweden during the 1930s (see Fisher, 1934, and Berg and Jonung, 1999; this
regime was quite successful in avoiding de¯ation).
Even if there are no current examples of price-level targeting regimes, price-
level targeting has been subject to an increasing interest in the monetary
policy literature. At the 1984 Jackson Hole Symposium, Hall (1984) argued for
price-level targeting. Several recent papers compare in¯ation targeting and
price-level targeting, some of which are collected in Bank of Canada (1994).
Some papers compare in¯ation and price-level targeting by simulating the
effect of postulated reaction functions. Other papers compare the properties of
postulated simple stochastic processes for in¯ation and the price level (see
Fischer, 1994). A frequent result, which has emerged as the conventional
wisdom, is that the choice between price-level targeting and in¯ation
targeting involves a tradeoff between low-frequency price-level variability on
the one hand and high-frequency in¯ation and output variability on the
other. Thus, price-level targeting has the advantage of reduced long-term
variability of the price level. This should be bene®cial for long-term nominal
contracts and intertemporal decisions, but it would come at the cost of
increased short-term variability of in¯ation and output. The intuition is
straightforward: in order to stabilise the price level under price-level targeting,
higher-than-average in¯ation must be succeeded by lower-than-average
in¯ation. This would seem to result in higher in¯ation variability than under
in¯ation targeting, since base drift is accepted in the latter case and higher-
than-average in¯ation need only be succeeded by average in¯ation. Via
nominal rigidities, the higher in¯ation variability would then seem to result in
higher output variability.5
However, this intuition may be misleadingly simple. In more realistic
models of in¯ation targeting and price-level targeting with more complicated
dynamics, the relative variability of in¯ation in the two regimes becomes an
open issue. As shown in Svensson (1997b, appendix), this is the case if there is
serial correlation in the deviation between the target variable and the target
level ± for instance, if the price level displays mean reversion towards the price-
level target under price-level targeting and in¯ation displays mean reversion
towards the in¯ation target under in¯ation targeting. Svensson (1999e) gives
an example where the absence of a commitment mechanism and at least
moderate persistence in the Phillips curve imply that in¯ation variability
becomes lower under price-level targeting than under in¯ation targeting,
without output variability becoming higher.6 For some empirical macro-

models (both small and large), reaction functions with responses of the
instrument to price level deviations from a price-level target lead to as good or
better overall performance (in terms of in¯ation and output variances) than
with responses to in¯ation deviation from in¯ation targets.7
I believe these results show that the relative properties of price-level
targeting and in¯ation targeting are far from settled. In particular, the
potential bene®ts from reduced long-term price-level variability and un-
certainty are not yet well understood. Still, I believe that low and stable
in¯ation may be a suf®ciently ambitious undertaking for central banks at
present. However, once central banks have mastered in¯ation targeting, in
perhaps another ®ve or ten years, it may be time to increase the ambitions and
consider price-level targeting. By then, research and experience may provide
better guidance about which regime is preferable.
The rest of the chapter will refer to `low in¯ation', corresponding to ( 2.4),
with possible additional terms in the loss function, rather than `price-level
stability', corresponding to (2.2). Reluctantly, I will occasionally refer to `low
in¯ation' as `price stability', without using quotation marks. Some of the
discussion below is applicable to both price-level stability and low in¯ation,
however.
The loss function

Is (2.4) an appropriate loss function for a monetary policy aimed at low
in¯ation? As reported below, there seems to be considerable agreement among
academics and central bankers that the appropriate loss function is instead of
the form
1
Lt t � 2 yt � yt 2 2:7
2
where yt is (log) output, yt is potential output, so that yt � yt is the output gap,
and > 0 is the relative weight on output-gap stabilisation. As in Svensson
(1999c) and (2000b), the case when 0 and only in¯ation enters the loss
function can be called strict in¯ation targeting, whereas the case when > 0
and the output gap (or concern about stability of the real economy in general)
enters the loss function can be called ¯exible in¯ation targeting.8
Whereas there may previously have been some controversy about whether
in¯ation targeting involves concern about real variability, represented by
output-gap variability and corresponding to the second term in (2.7), there is
now considerable agreement in the literature that this is indeed the case.
In¯ation-targeting central banks are not what King (1997) called an ìn¯ation
nutter'. For instance, Fischer (1996), King (1996), Taylor (1996) and Svensson
(1996) in Federal Reserve Bank of Kansas City (1996), all discuss in¯ation
targeting with reference to a loss function of the form (2.7) with > 0. As
shown in Svensson (1997a) and Ball (1997), concern about output-gap
stability translates into a more gradualist policy. Thus, if in¯ation moves away
from the in¯ation target, it is brought back to target more gradually.
Equivalently, in¯ation-targeting central banks lengthen their horizon and
aim at meeting the in¯ation target further in the future. As further discussed in
Svensson (1999c), concerns about output-gap stability, simple forms of model
uncertainty and interest rate smoothing all have similar effects under in¯ation
targeting ± namely, a more gradualist policy. The Sveriges Riksbank has
explicitly expressed very similar views.9 The Chancellor's remit to Bank of the
England (HM Treasury, 1997) mentions ùndesired volatility of output'.10 The
Minutes from Bank of England's Monetary Policy Committee (Bank of
England, 1999) are also explicit about stabilising the output gap.11 Several
contributions and discussions by central bankers and academics in Lowe
(1997) express similar views. Ball (1999) and Svensson (1998b) give examples
of a gradualist approach of the Reserve Bank of New Zealand. Indeed, a quote
from the ECB (European Central Bank, 1999 p. 47) also gives some support for
an interpretation with > 0, as well as some weight on minimizing interest
rate variability:
a medium-term orientation of monetary policy is important in order to

permit a gradualist and measured response [to some threats to price
stability]. Such a central bank response will not introduce unnecessary and
possibly self-sustaining uncertainty into short-term interest rates or the real
economy
Thus, it is seems non-controversial that real world in¯ation targeting is

actually ¯exible in¯ation targeting, corresponding to > 0 in (2.7).
The loss function (2.7) highlights an asymmetry between in¯ation and
output under in¯ation targeting. There is both a level goal and a stability goal
for in¯ation, and the level goal ± that is, the in¯ation target ± is subject to
choice. For output, there is only a stability goal and no level goal. Or, to put it
differently, the level goal is not subject to choice; it is given by potential
output. Therefore, I believe it is appropriate to label minimising (2.7) as
`(¯exible) in¯ation targeting' rather than ìn¯ation-and-output-gap targeting',
especially since the label is already used for the monetary policy regimes in
New Zealand, Canada, United Kingdom, Sweden and Australia.
What index and which level?

Which price index would be most appropriate? Stabilising the CPI should
simplify consumers' economic calculations and decisions. The CPI has the
advantage of being easily understood, frequently published, published by
authorities separate from central banks and very rarely revised. Interest-related
costs cause well known problems with the CPI, though: an interest rate
increase to lower in¯ation has a perverse short-term effect in increasing
in¯ation. It makes sense to disregard this short-term effect in monetary policy
decisions, but it still presents a pedagogical problem in the central bank's

communication with the general public. To avoid this problem, the Bank of
England and the Reserve Bank of New Zealand have in¯ation targets de®ned in
terms of CPIX (RPIX in Britain), the CPI less interest-related costs.12 The
Eurosystem has also de®ned price stability in terms of the HICP (Harmonised
Index of Consumer Prices), which excludes interest costs. Furthermore,
changes in indirect taxes and subsidies can have considerable short-run effects
on the CPI. Different measures of underlying in¯ation ± core in¯ation ± try to
eliminate such effects. Eliminating components over which monetary policy
has little or no in¯uence serves to avoid misleading impressions of the degree
of control. The disadvantage with subtracting too many components from the
index used for the in¯ation target is that the index becomes more remote from
what matters to consumers and less transparent to the general public. It may
also be dif®cult to compute in a well de®ned and transparent way. Opinions
generally differ on what components to deduct from the CPI. My own
view is that deducting interest-related costs and using CPIX, together with
transparent explanation of index movements caused by changes in indirect
taxes and subsidies, is an appropriate compromise.
What level of the in¯ation target is appropriate? Although zero in¯ation
would seem to be a natural focal point, all countries with in¯ation targets have
selected positive in¯ation targets. The in¯ation targets (point targets or mid-
points of the target range) ranging between 1.5 per cent (per year) in New
Zealand, 2 per cent in Canada, Sweden and Finland (before joining EMU), and
2.5 per cent in the United Kingdom and Australia (the Reserve Bank of
Australia has an in¯ation target in the 2±3 per cent range for average in¯ation
over an unspeci®ed business cycle). The Bundesbank had a 2 per cent in¯ation
target for many years (called ùnavoidable in¯ation', `price norm', or
`medium-term price assumption'). During 1997 and 1998, it was lowered to
1.5±2 per cent (which could perhaps be translated into a point in¯ation target
of 1.75 per cent). The EMI ± (European Monetary Institute, 1997) de®ned price
stability as 0±2 per cent. The Eurosystem has announced ànnual increases in
the HICP below 2 per cent' as its de®nition of price stability, which has been
interpreted as intervals 0±2 per cent or 1±2 per cent; the Eurosystem used a
point in¯ation target of 1.5 per cent in constructing its reference value for
money growth. The Eurosystem's de®nition of price stability is further
scrutinised on page 88.
That the in¯ation target exceeds zero can be motivated by measurement
bias, non-negative nominal interest rates and possible downward nominal
price and wage rigidities.13 2 per cent is the borderline in Akerlof, Dickens and
Perry (1996), who study the effects of downward rigidity of nominal wages. 1
per cent is the borderline in Orphanides and Wieland (1998), who examine
the consequences of non-negative nominal interest rates. These studies
indicate that in¯ation targets below those borderlines risk reducing average
output or increasing average unemployment.14 Altogether, announcing an
explicit in¯ation target (a point target or a range) may be more important than
whether the target (the mid-point of the range) is 1.5, 2 or 2.5 per cent.
A symmetric in¯ation target implies that in¯ation below the target is
considered equally bad as in¯ation the same distance above the target (which
is the case if in¯ation targeting is represented by a symmetric loss function like
(2.7)). This would seem to be a precondition for in¯ation expectations being
focused on the in¯ation target. A point target with or without a tolerance
interval would, from this point of view, be better than just a range. A range
would, in turn, be better than an asymmetric formulation like `below 2 per
cent'. These aspects may be particularly important when persistent de¯ation is
a possibility, of which recent developments in Japan remind us. A symmetric
in¯ation target would seem to be the best defence against persistent de¯ation
and against the appearance of de¯ationary expectations.
Interestingly, a price-level target may have special advantages relative to an
in¯ation target in avoiding persistent de¯ation, since an unanticipated
de¯ation which makes the price level fall below the price-level target will, if
the price-level target is credible, result in increased in¯ation expectations that
will, in themselves, reduce the real interest rate and stabilise the economy.15
3 Maintaining price stability
The basic problem of maintaining price stability is thus to set the monetary
policy instrument (or instruments) so as to minimise the intertemporal loss
function (2.1) with the period loss function (2.7), subject to current
information about the current and future state of the economy and the
transmission mechanism.
The transmission mechanism is taken to be represented by a linear model in
state±space form

Xt1 X ut1
A t Bit 2:8
xt1jt xt 0
where Xt t ; yt ; yt ; ::; 10 (where 0 denotes transpose) is a column vector of nX
predetermined variables (also called state variables), xt is a column vector of nx
forward-looking variables (also called non-predetermined variables), it is a column
vector of ni central bank instruments (also called control variables), ut1 is a
column vector of nX exogenous iid shocks with zero means and a constant
covariance matrix uu and A and B are matrices of appropriate dimensions. The
predetermined variables include in¯ation, output, potential output and other
variables. I use the convention that the last element of the vector of
predetermined variables is unity. This is a convenient way of allowing non-zero
means of the variables; the last column of A is then a function of these means.
Although the framework is general enough for handling multiple monetary
policy instruments I will, realistically, assume that there is only one
instrument (ni 1) and take that instrument to be a short nominal interest

rate (for instance, an overnight interest rate or a one- or two-week repurchase
rate).
The expression xt1jt denotes Et xt1 , the expectation of xt1 conditional
upon all information available in period t, including any information about
the state of the economy and the model of the economy. The forward-looking
variables include asset prices, like exchange rates and interest rates of longer
maturity than the instrument, and other variables partially or fully
determined by the expectations of future variables.
Thus, at this stage I assume that there are no non-linearities in the
transmission mechanism (or that shocks and deviations from a steady state are
moderate so a linear approximation is acceptable). Furthermore, I make the
assumption that the model is known, that the central bank and the private
sector have the same information and that the predetermined and forward-
looking variables in period t are observable in period t. I will discuss
generalisations of those assumptions below.
A more general representation of the monetary policy loss function is to let
Yt denote a column vector of nY target variables, given by

Xt
Yt C Ci it
xt
where C and Ci are matrices of appropriate dimension. Let Y denote the

column vector of nY target levels, and let the period loss function be
Lt Yt � Y 0 WYt � Y
where W is a positive-semide®nite weight matrix. The period loss function

(2.7.) is a special case of this more general loss function, where the target
variables are given by Yt t ; yt � yt 0 , the target levels by Y ; 0 and the
weight matrix W is a diagonal matrix with the diagonal 1=2; =2.
Given this representation of the loss function and the transmission
mechanism, the problem is now to ®nd the principles for setting the
instrument it in each period t. I will consider two such main principles, ®rst
what can be called `commitment to an instrument rule' or ìnterest-rate
targeting' and then `forecast targeting' (p. 73). A third principle, ìntermediate-
variable targeting', especially monetary targeting, is brie¯y considered on
p. 86.
Commitment to a simple instrument rule: interest rate targeting

Make the unrealistic assumption that the central bank can commit, once and
for all, in period t 0 to a particular reaction function for all future periods.
Furthermore, assume that the model (2.8) is known, that the predetermined
and forward-looking variables are observable in each period, and that X0 is
given. Under these assumptions, it is possible to ®nd the optimal reaction
function under commitment that minimises (2.1) in period 0 (see Backus and
Drif®ll, 1986; Currie and Levine, 1993; and Soderlind, 1998). This reaction
function will be a linear function of the predetermined variables and the
predetermined Lagrange multipliers (shadow prices) of the forward-looking
variables,
it f Xt 't 2:9
for t 0 and X0 given, where f and ' are row vectors with nX and nx elements
(called response coef®cients, or reaction coef®cients), respectively. Further-
more, the multipliers t ful®l
t1 M21 Xt M22 t 2:10
for t 0 and 0 0, where M21 and M22 are matrices of appropriate

dimension. It follows from (2.9) and (2.10) that the optimal reaction function
under commitment can be written as a distributed lag of past predetermined
variables,
X
t
�1
it f Xt ' M22 M21 Xt� 2:11
1
If there are no forward-looking variables, there is no distinction between

commitment and discretion. Furthermore, the optimal reaction function is a
linear function of the current predetermined variables only,
it f Xt 2:12
Even when there are forward-looking variables, many papers consider the
optimal reaction function under commitment over the class of reaction
functions (2.12) of the current predetermined variables only (mostly without
notifying the reader that this is a restriction).
The optimal reaction function under commitment is normally a function of
all the predetermined variables (and the lagged predetermined variables) and
is, in this sense, a rather complex construction. Consider also the class of
simple reaction functions, the class of linear reaction functions restricted to
being `simple' in the sense of having few arguments (for instance, some of the
elements of vector f and all the elements of vector ' are restricted to zero). A
typical simple reaction function is the much-discussed Taylor rule, where the
instrument responds only to current or lagged in¯ation and the output gap.
Let the optimal simple reaction function under commitment be the reaction
function in a particular class of simple reaction functions that minimises (2.1)
in period 0, given X0 :
An optimal reaction function under commitment is likely to be too
complex, in the sense of involving speci®c responses to a large number of
predetermined variables, to be veri®able. Therefore it is dif®cult to conceive of
a commitment of the central bank to this reaction function. A simple reaction

function is easier to verify. Therefore, in principle we can conceive of a
commitment to a simple reaction function, a commitment to a simple
instrument rule.
Such a commitment could also be expressed as a targeting rule, more
precisely a commitment to the particular loss function corresponding to
ìnterest rate targeting'. Then, for a particular simple reaction function f , a
time-varying interest rate target, it , is de®ned as
it f Xt 2:13
Then, instead of the period loss function (2.7), the central bank is committed
to the new period loss function
1
Lt it � it 2 2:14
2
Clearly, a trivial ®rst-order condition for minimising (2.14) is given by16
it it 2:15
Thus, (2.15) can either be interpreted as a targeting rule, a ®rst-order condition

resulting from the commitment to the particular loss function (2.14) with the
interest target (2.13), or it can be interpreted directly as an instrument rule, a
prescribed rule for setting the instrument as a function of observed variables.
As an example, we can consider a Taylor-type reaction function with
smoothing, which corresponds to an interest rate target given by
it 1 � p r g t � gy yt � yt pit�1
where r is the average real interest rate, g and gy are the long-run response
coef®cients and 0 < 1 is a smoothing parameter.
Furthermore, we realise that under this commitment to a simple policy rule,
the central bank need no longer be forward-looking. It need only be forward-
looking once and for all in period 0, when it decides to which simple reaction
function it will commit. After that, it need never be forward-looking; to set the
instrument according to the prescribed reaction function, or it simply needs to
minimise the period loss function each period with the prescribed interest rate
target.
Although most of the current and previous discussion of monetary policy
rules is in terms of commitment to alternative instrument rules (see, for
instance, McCallum, 1997, and the contributions in Bryant, Hooper and
Mann, 1993 and Taylor, 1999), I do not believe that a commitment to an
instrument rule is either a practical or desirable way of maintaining price
stability, for several reasons.
First, there are overwhelming practical dif®culties in deciding once and for
all which instrument rule to follow. The optimal reaction function will involve
speci®c responses to a large number of (current and lagged) information

variables and is therefore unlikely to be veri®able. Furthermore, results by
Levin, Wieland and Williams (1998) indicate that the optimal reaction
function (in their case with the restriction that ' is zero) is quite sensitive to
the model. This is problematic, since the model is, in practice, not precisely
known. A simple reaction function may be more robust, in the sense of
performing reasonably well in different models. This is an idea promoted and
examined in several papers by McCallum and recently restated in McCallum
(1997). The results of Levin, Wieland and Williams also indicate that a simple
reaction function may be quite robust in this sense. On the other hand, as
shown by Currie and Levine (1992), the optimal simple reaction function does
not only depend on the model and the loss function but also on the stochastic
properties of the shocks and the initial state of the economy, X0 , so that the
performance of simple rules generally depends on these stochastic properties
(certainty-equivalence does not hold for simple reaction functions in linear
models with quadratic loss functions, in contrast to the case for the optimal
unrestricted reaction function).17
Second, a commitment to an instrument rule does not leave any room for
judgemental adjustments and extra-model information. As argued further
below, the use of judgemental adjustments and extra-model information is
both unavoidable in practice and desirable in principle. Also, there is no room
for revision of the instrument rule, when new information results in a revision
of the model. By disregarding such information, a commitment to an
instrument rule would be inef®cient.
Third, although a commitment to a complex instrument rule also seems
inconceivable in principle, since it will hardly be veri®able, a commitment to a
simple instrument rule is, in principle, feasible ± for instance, by an interest
rate targeting regime as above. Still, such a commitment is unheard of in the
history of monetary policy, for obvious reasons. It would involve committing
the decision making body of the central bank to reacting in a prescribed way to
prescribed information. Monetary policy could be delegated to the staff, or
even to a computer. It would be completely static and not forward-looking.
Such a degradation of the decision making process would naturally be strongly
resisted by any central bank and, I believe, arguments about its inef®ciency
would easily convince legislators to reject it as well. In practice, there is
therefore no commitment mechanism that commits the decision making
body to reacting in a prescribed way to prescribed information. In practice,
decision making under considerable discretion is unavoidable, and nothing
prevents the decision making body from reconsidering their decisions more or
less from scratch, without being bound by previous decisions and commit-
ments. As Blinder (1998, p. 49) puts it, Ènlightened discretion is the rule'.
Fourth, in the absence of a commitment mechanism, a prescribed simple
instrument rule would not be incentive-compatible. There would be frequent
incentives to deviate, often for very good reasons, due to new, unforeseen,
information (stock market crash, Asian crisis, Brazil ¯oating) and correspond-
ing sound judgemental adjustment.
Although alternative instrument rules can serve as informative guidelines
(see, for instance, the contributions in Taylor, 1999 ± or, with regard to the
performance of a Taylor rule for the Eurosystem ± Gerlach and Schnabel, 1998,
Peersman and Smets, 1998 and Taylor, 1998) and decisions ex post may
sometimes be similar to those prescribed by the simple instrument rules, they
are not a substitute for a decision making procedure for the central bank.
Interest rate targeting for the Eurosystem was indeed rejected by the European
Monetary Institute (EMI), the predecessor of the European Central Bank (ECB),
in European Monetary Institute (1997, p. 1) (with, arguably, not the most
exhaustive argument):
[T]he use of an interest rate as an intermediate target is not considered

appropriate given dif®culties in identifying the equilibrium real interest
rate which would be consistent with price stability.
Indeed, instead of having a decision making procedure and being forward-

looking only once and for all, at the time of a commitment to a simple rule, the
central bank needs to have a continuous decision making procedure and be
continuously forward-looking. To quote Greenspan (1994, p. 244):
Implicit in any monetary policy action or inaction, is an expectation of

how the future will unfold, that is, a forecast.
The belief that some formal set of rules for policy implementation can
effectively eliminate that problem is, in my judgement, an illusion. There is
no way to avoid making a forecast, explicitly or implicitly.18
Therefore, I now turn to a practical and realistic, and indeed already

practised, way of maintaining price stability, namely by way of `forecast
targeting'.19
Forecast targeting
As a background, recall that, with a quadratic loss function and a linear model,
with the assumption of a known model and only additive uncertainty,
certainty-equivalence applies. The problem of minimising the loss function
can be separated into a deterministic problem involving conditional forecasts,
the conditional means of current and future variables and a stochastic problem
involving deviations between realised outcomes and conditional forecasts.
The solution to the deterministic problem is the same as to the stochastic
problem (see Chow, 1975, for models with only predetermined variables and
Currie and Levine, 1993, for models with forward-looking variables as well).
Thus, the discussion can focus on the deterministic problem involving
conditional forecasts.
For any variable t , let tjt for 0 denote the expectation Et t given
information in a ®xed period t. The information in period t includes the
information available about the state of the economy as well as about the
model, (2.8).20 Let jt denote the future path tjt ; t1jt ; t2jt ; ::: Consider the set
I t of given paths ijt itjt ; it1jt ; ::: of instrument settings, for which there exist
bounded paths jt and yjt � yjt of future in¯ation and output gaps,
respectively. For each ijt 2 I t , let jt ijt denote the corresponding path for
variable and y (yjt is taken to be exogenous), and call it the corresponding
conditional forecast (conditional on information in period t, ijt and the model
(2.8)). Let
n o
YT jt ijt ; yjt ijt � yjt jijt 2 I t
denote the set of feasible conditional forecasts of the target variables.

Constructing conditional forecasts in a backward-looking model (that is, a
model without forward-looking variables) is straightforward. Constructing
such forecasts in a forward-looking model raises some speci®c dif®culties,
which are explained and resolved in Svensson (1998a, Appendix A).21
Due to the certainty-equivalence, the stochastic optimisation problem of
minimising the expected intertemporal loss function (2.1) over future random
target variables in (2.7), subject to (2.8), is equivalent to the deterministic
problem of minimizing the deterministic intertemporal loss function
X
1
L~ t jt 2:16
0
with the deterministic period loss function

1
L~ t jt tjt � 2 ytjt � yt
2
jt 2:17
2

(where the stochastic t and yt � yt have been replaced by the

deterministic tjt and ytjt � yt ) subject to
jt; yjt � yjt 2 Y t 2:18
Thus, the problem of the central bank is to choose the path ijt ; such that the
resulting jt and yjt minimise (2.1) with (2.17). The ®rst element of ijt , itjt , is
then the appropriate instrument setting for period t; it . If there is no new
relevant information in period t 1, the instrument setting in period t will be
the second element in ijt . If there is new relevant information, that
information is used for solving the problem again in period t 1.22
This procedure thus involves making conditional forecasts of in¯ation,
output and the output gap for alternative interest rate paths, using all relevant
information about the current and future state of the economy and the
transmission mechanism. It involves making consistent assumptions about
exogenous and endogenous variables (for instance, that exchange rates and
interest rates are consistent with appropriate parity conditions). As discussed
further below, it also allows judgemental adjustments and extra-model
information (p. 77), as well as partially observable states of the economy
(p. 79). As discussed on p. 82, forecast targeting can even be adapted to take
non-linearities and model uncertainty into account, which both result in non-
additive uncertainty.
The procedure requires estimates of policy multipliers, the effects on the
conditional forecasts of changes in the instrument. The policy multipliers are
easily calculated in a simpli®ed model with only predetermined variables
Xt1 AXt Bit ut1 2:19
(see Svensson 1998a, Appendix A, for a case with forward-looking variables).

The conditional forecast for Xtjt then ful®lls
X
�1
Xtjt AXt�1jt Bit�1jt A Xtjt A�1�s Bitsjt 2:20
s0
for 1, so that the policy multipliers dXtjt =ditsjt ; 0 s � 1, are given by
dXtjt
A�1�s B 2:21
ditsjt
Optimality criterion
What is the criterion for having found an optimal interest rate path and
corresponding conditional forecasts of in¯ation and the output gap? One
criterion can be formulated as follows. Suppose the central bank staff have
constructed a potential optimal combination of an interest-rate path and such
conditional forecasts. Consider a change dijt ditjt ; dit1jt ; ::: in the interest-
rate path ijt . This will result in changes dXjt 0; dXt1jt ; dXt2jt ; ::: in the
predetermined variables, given by
X
�1
dXtjt
dXtjt ditsjt
s0
dits
Let djt and dyjt denote the corresponding changes in the in¯ation and output
forecasts (the output-gap forecast yjt is taken to be exogenous). A necessary
condition for optimality is then that the corresponding change in the
intertemporal loss function is non-negative ± that is
X
1 X
1
d Ltjt tjt � dtjt
0 0

ytjt � ytjt dytjt 0 2:22
Given jt , yjt , yjt d;jt and dyjt ; as well as , and ; this expression is easily
checked, and a relatively easy way of making pairwise comparisons of
alternative interest rate paths and conditional forecasts. For instance, a delay
in an interest rate increase can be compared with an immediate increase, a
small increase now can be compared with a larger increase two quarters later,
etc.
Within the simple model in Svensson (1997a, 1999c), I have shown that the
®rst-order conditions for an optimum can be expressed as particularly simple
targeting rules (in the form of equations for the conditional forecasts of the
target variables). For instance, the in¯ation gap t1jt � and the output

gap ytjt � ytjt should be of the opposite signs and related according to
y c
t1jt � �
ytjt � ytjt ; 1
1 � c
where the coef®cient y is the sensitivity of the change in in¯ation to the

output gap and the coef®cient c is an increasing function of that ful®ls
0 c < 1, c0 0 and c1 1. Alternatively, the targeting rule can be
expressed as the in¯ation forecast approaching the in¯ation target at a
constant rate
t1jt � ctjt � ; 1
In practice, the decision making body may get far by just visually examining
alternative in¯ation and output-gap forecasts and then choosing the one that
is the best compromise between hitting the in¯ation target at an appropriate
horizon and avoiding output-gap stability. If done in a consistent way, this
will be equivalent to minimising the intertemporal loss function. Compared
to many other intertemporal decision problems that households, ®rms and
investors solve one way or another (usually without the assistance of a
substantial staff of economics PhDs), this particular decision problem does not
seem to be overly complicated or dif®cult.
Instrument assumptions
In the above discussion, the problem is to ®nd the appropriate interest rate
path ijt , which requires constructing conditional forecasts for exogenous
interest rate paths.
Forecasts for unchanged interest rates, where the interest rate path ful®ls
itjt it�1; 0
can be used as indicating `risks to price stability'. They are used by Bank of
England and Sveriges Riksbank to motivate changes in the interest rate and
their direction (see p. 81f.)
Conditional forecasts can also be constructed for given reaction functions,
in which case the interest rate path is endogenous and ful®ls
itjt f Xtjt
for a given f .23 Some central banks, notably the Bank of Canada and the
Reserve Bank of New Zealand, construct forecasts conditional on reaction
functions involving what Rudebusch and Svensson (1999a) call `responding to
model-consistent forecasts', and what Batini and Haldane (1999) call `forecast-
based' reaction functions. This implies an interest rate path ful®lling
itjt f Xtjt gXtT jt;
where T > 0 is the forecast horizon (typically some six±eight quarters), and
sometimes f 0:
Among reaction functions involving responses to forecasts, it would seem
more natural that the forecast responded to is one for unchanged interest rates.
Indeed, under strict in¯ation targeting, the optimal instrument adjustment is
proportional to the deviation of the conditional forecast for unchanged interest
rate from the in¯ation target, as demonstrated in Svensson (1999d).24
Judgemental adjustments
A major advantage of forecast targeting relative to commitment to an
instrument rule is that it allows a systematic and disciplined way of
incorporating judgemental adjustment and extra-model information, by
`®ltering information through the forecasts'.
Given that every model of the transmission mechanism is an abstraction
and a simpli®cation, and given that there is considerable uncertainty about
the details of the transmission mechanism, monetary policy will never, it
seems, be able to rely on models and the information entering models alone.
There will always be an important role for additional extra-model information
and judgemental adjustments of the instrument. Never using such informa-
tion and judgement would neither be ef®cient nor incentive-compatible for
the decision making body of the central bank. At the same time, such
information and informal adjustment opens up monetary policy to arbitrari-
ness and potential abuse. Forecast targeting allows some system and discipline
in the use of extra-model information and judgemental adjustments.
Judgemental adjustments can be made in several ways in the framework
outlined here. One is in the form of adjustments of the coef®cients of the
model. This means that the coef®cients of matrices A and B become time-
varying, Atjt and Btjt , which is easily incorporated when constructing the
forecasts (as long as the time-variation is deterministic; see p. 84 on non-
additive uncertainty).
Another kind of judgemental adjustments consists of simple additive shifts
in the forecasts. For the model with only predetermined variables, this means
that the model in period t can be represented as
Xt1jt AXtjt Bitjt jtjt

where the column vector jtjt corresponds to additive judgemental adjust-

ments to Xt in period t. Suppose, as above, that the last element in Xt is
unity, in order handle non-zero means. Add the additive judgemental
adjustment jtjt to the last column of the matrix A to form the matrix
Atjt . Then, the system can be written
Xt1jt Atjt Xtjt Bitjt
formally as a model with time-dependent coef®cients.25

Only if a particular piece of information can be convincingly shown to affect
the conditional forecasts at relevant horizons does it warrant a change in the
instrument. It is not correct to adjust the instrument without `®ltering the
information through the forecasts'. Thus, targeting rules and forecast targeting
bring some system and discipline to the use of judgments and extra-model
information, and provide some protection against the arbitrariness in use and
temptations of abuse that might easily arise. This is, I believe, one aspect of
what Bernanke and Mishkin (1997) call `constrained discretion' (although
they are not explicit about the role of forecast targeting).26
Mean, median or mode forecasts?
Is it the mean, the median or the mode forecast that is relevant in forecast
targeting? Under a linear model with additive uncertainty and a quadratic loss
function, the previous discussion have demonstrated that certainty-equiva-
lence holds and that it is the mean forecast that is relevant. When would
median or mode forecasts be relevant? Vickers (1998) and Wallis (1999) have
provided systematic discussions of this issue. Under the maintained assump-
tion of a linear model with additive uncertainty and objectives corresponding
to minimum expected loss, the nature of the loss function determines which
forecast is relevant. Thus, with a quadratic loss function, it is the mean
forecast. Under a linear loss function (that is, linear in the deviation from
target, V-shaped), it is the median forecast. Under an àll or nothing' loss
function with extreme favourable weight on the target level, it is the mode
forecast. Finally, under a loss function assigning a constant loss outside a
tolerance interval and a zero loss inside, the forecast should maximise the
probability of being inside the interval, which implies that the upper and
lower bounds should have the same probability density.27
The mean, median and mode alternatives differ only when the probability
distribution is asymmetric. When they differ, unless certainty-equivalence is
explicitly assumed not to hold, it seems that the mean forecast should still
have prominence and never be excluded, also when the mode and/or the
median is reported. Still, both the Bank of England and the Sveriges Riksbank
report the mode forecast as their central forecast, in spite of the mode being
associated with a relatively bizarre loss function, the all-or-nothing one.28
The Bank of England and the Sveriges Riksbank, however, aim to illustrate
not only a point forecast but the probability distribution of the forecasted
variable, so as to indicate the uncertainty of the forecast, as well as the `balance

of risks', ± that is, whether the distribution is symmetric or not. Thus, the Bank
of England produces its famous fan chart, which can be interpreted as showing
iso-level contours for the density function of the probability distribution, where
each shade of colour (red for in¯ation, green for the output forecast) encloses a
given fraction of the probability mass (see Britton, Fisher and Whitley, 1998).
Thus, the fan chart displays the density function in the same way as a contour
map illustrates a hill (with the difference that maps usually have contours with
equal vertical distance whereas the fan chart has equal increments of the
enclosed probability mass). Furthermore, the mode will be at the centre of the
narrow central band, the deepest shade of colour which denotes the most likely
outcomes, the highest portion of the probability density. As noted by Wallis
(1999), this means that the Bank of England uses non-central prediction (or
con®dence) intervals, which are different from standard central prediction
intervals in that they do not have the same probability mass above and below
the intervals (when the distribution is asymmetric).
The Sveriges Riksbank instead reports the mode together with standard
central prediction intervals, with the same probability above and below the
prediction interval (see Blix and Sellin, 1998). As a consequence, for an
asymmetric distribution, the mode will not be at the centre of the most narrow
con®dence interval (instead, the centre of the most narrow con®dence interval
is the median).
As long as these alternative graphs are clearly understood as different ways of
illustrating the whole distribution, and policy is based on the whole distribution,
the ways of illustration need not have any effect on the policy. Still, with the ways
of illustration that these central banks have chosen, I believe that most observers
are led to focus on the mode, when the mean arguably in many cases is a more
appropriate focus. Therefore I believe that it might be better to plot the mean
rather than the mode, together with the different con®dence intervals. I believe
this might be better also under distribution forecast targeting, to be discussed on
p. 82, when the whole distribution matters.29
The role of indicators

So far, the maintained assumption in the discussion has been that the central
bank can directly observe the state of the economy, more precisely, observe
the predetermined and forward-looking variables, Xt and xt in period t. In
order to discuss the role of indicators, I now assume that the predetermined
and forward-looking variables are not necessarily observable. Consequently,
introduce a vector of nZ observable variables, indicators, Zt . The indicators
depend on the predetermined and forward-looking variables and the
instrument according to

Xt
Zt D Di it vt
xt
where D and Di are matrices of appropriate dimension and vt is a vector of nZ

iid shocks with zero means and constant covariance matrix vv : These shocks
may be interpreted as measurement errors. Assume that the central bank's
information in period t is represented by the information set

It Zt�; 0; A; B; C; Ci ; D; Di ; uu; vv g
That is, in period t, the central bank knows the current and past indicators, in
addition to the model and the stochastic properties of the shocks.
This formulation allows some variables to be directly observable, some to be
observed with measurement error and some to be completely unobservable.
The discussion in previous sections, the full information case, corresponds to
the special case Zt X0t ; x0t ; i0t 0 .
For simplicity, assume that there are no forward-looking variables, so that
the model is (2.19), and let the indicators depend on the predetermined
variables according to
Zt DXt vt 2:23
This setup is examined in more detail in Chow (1975), Tinsley (1975) and
LeRoy and Waud (1977).30 The case of partial information with forward-
looking variables is examined in Pearlman, Currie and Levine (1986),
Pearlman (1992), Aoki (1998) and Svensson and Woodford (2000). The
present discussion follows Svensson and Woodford (2000) (although without
forward-looking variables).
Under these assumptions, certainty-equivalence continues to hold. In
period t, the central bank needs to form the estimate Xtjt of the predetermined
variables in order to construct its conditional forecasts and set its instrument.
The optimal estimate is given by a Kalman ®lter, with the updating equation
Xtjt Xtjt�1 KZt � Ztjt�1 2:24
where the matrix K is the Kalman gain matrix.31

The elements of the gain matrix give the optimal weights on the indicators in
estimating the predetermined variables. Using (2.23), the updating equation can
be written
Xtjt I � KD Xtjt�1 KZt 2:25
so the matrices K and I � KD give the weights on the indicators Zt and the
prior information Xtjt�1 . If D I and vv 0, Zt coincides with the
predetermined variables; then K I and all weight is on the indicators and
none on the prior information. Generally, a row in K gives the optimal
weights on the indicators in estimating the corresponding predetermined
variable. A column in K gives the weights given to the corresponding
indicator in estimating the different predetermined variables. Assume that a
particular indicator, Zjt , is equal to one of the predetermined variables, Xkt ,
plus a measurement error, vjt ;
Zjt Xkt vjt
If the variance of the measurement error approaches zero, all the elements in
row k of K approaches zero except element j, which approaches unity. Thus, all
the weight in estimating Xkt falls on Zjt . If the variance of the measurement
error goes to in®nity, Zjt becomes a useless indicator. The elements in column j
of K then all become zero. The indicator Zjt gets zero weight in estimating the
predetermined variables.
Assume, for simplicity, that forecast targeting has resulted in a reaction
function f in the past. That is, it� f Xt�jt� for 1 t. Then, by (2.9),
Xj�1 A Bf X�1j�1
for 1 t. Using this in (2.25), we can write the updating equation as a

distributed lag of past indicators,
Xtjt 1 � KDA Bf Xt�1jt�1 KZt

1 � KDA Bf t X0j0
X
t�1
1 � KDA Bf KZt�
0
This gives the weight on indicators Zt� as 1 � KDA Bf K for 0.
These equations illustrate the gradual updating of the estimate of the
predetermined variables. We can summarise the effects of the indicators in
period t on the forecasts Xtjt in terms of ìndictor multipliers,' dXtjt =dZt ,
given by
dXtjt dXtjt dXtjt
A K
dZt dXtjt dZt
Thus, the indicator multiplier is the product of the effect of the estimate of the
current state of the forecast, dXtjt =dXtjt , which by (2.20) is given by A , and
the effect of the indicators on the estimate of the current state, with by (2.24) is
given by the gain matrix K. It follows that an indicator will affect the
instrument setting via affecting the current state of the economy, then the
forecasts, and ®nally the instrument. Schematically,
Zt ! Xtjt ! Xtjt ! it
In particular, the weights on any given indicator depend exclusively on its

power in predicting future in¯ation and output gap. Monetary aggregates, for
instance, have no special role beyond that, and any weight on monetary
aggregates will exclusively depend on its predictive performance.
An indicator of `risks to price stability'
An indicator of what the Eurosystem calls `risks to price stability' ± that is,
in¯ationary or de¯ationary pressure, or risks of overshooting or undershooting
the in¯ation target, ± should be useful when discussing monetary policy that
aims to maintain price stability. What requirements should such an indicator
ful®l? It would seem that, ®rst, it should signal in which direction and to what
extent the in¯ation target will be missed if policy is not adjusted. Second, it
should signal in which direction and to what extent the instrument should be
adjusted. Finally, it should be intuitive and easy to understand, so that it can
be used to communicate with the public and explain why an instrument
change is warranted or not.
The Eurosystem has put forward a money growth indicator, namely, the
deviation between current M3 growth and a reference value as an indicator of
risks to price stability - indeed, the ®rst of the `two pillars' of its monetary
strategy. As discussed in Svensson (1999d), such a money growth indicator
seems quite unsuitable for this purpose, since it is largely just a noisy indicator
of the deviation of current in¯ation from the in¯ation target (which deviation
can be more easily observed directly).32 Instead, the obvious candidate is a
conditional in¯ation forecast, conditional upon unchanged monetary policy
in the form of an unchanged interest rate. That is, it is constructed for the
interest rate path that ful®ls itjt it�1 , 0. Constructing this in¯ation
forecast is straightforward in a model without predetermined variables, as is
apparent from (2.20). It is somewhat more complicated in a model with
forward-looking variables, as shown in Svensson (1998a Appendix A).33 This
indicator signals whether and in which direction the in¯ation target is likely to
be missed, if policy is not adjusted, and it thereby also signals in which
direction the instrument needs to be adjusted.
The Bank of England and the Sveriges Riksbank, in their quarterly In¯ation
Reports, use conditional in¯ation forecasts for unchanged interest rates as their
main vehicle for motivating why the instruments need to be adjusted or not.
Complications to mean forecast targeting and generalisation to

distribution forecast targeting
Non-linearities and non-additive uncertainty in the model are two complica-
tions to forecast targeting. These complications both imply that the certainty-
equivalence underlying conditional mean forecasts no longer holds. These
two complications are discussed on p. 82 and p. 84.34 A solution to the
complications is to move from conditional mean forecast targeting to
conditional distribution forecast targeting, which is discussed on p. 85.
Non-linearities
So far, the maintained assumption has been that the model is linear. Sources of
non-linearity that have been discussed in the literature include non-linear
Phillips curves (see Debelle and Laxton 1997; Gordon, 1997; Isard and Laxton,
1996), non-negativity of nominal interest rates and downward nominal rigidity
of prices and/or wages (see references on pp. 66±8). Suppose now that the model
is non-linear. Assume that the model remains known, that there are no
forward-looking variables, that the predetermined variables are observable,

and that the model can be written as
Xt1 MXt ; it ; ut1 2:26
where M is a non-linear function. This has two consequences. First, the
conditional mean forecasts, Xtjt , are now non-linear functions of the interest
rate path, ijt , the current state of the economy, Xt , and the covariances of the
shocks, uu . Second, the policy multipliers, dXt =ditsjt ; will be stochastic,
and the `forecast' policy multipliers, dXtjt =ditsjt ; will be endogenous and not
constant. Then, certainty-equivalence no longer applies, and using the period
loss function (2.17) is no longer equivalent to using (2.7).
The reason why certainty-equivalence no longer holds can be demonstrated
with reference to the optimality criterion (2.22). Consider the change in the
intertemporal loss function (2.1.) with (2.7.) of a change dijt for the optimal
instrument path, which implies the optimality criterion
X
1
0 dEt Lt
0
X
1
1
d Et t � 2 yt � yt 2
0
2
X
1

Et t � dt Et yt � yt

dyt
0
X
1 n o
tjt � dtjt ytjt � ytjt

dytjt
0
X
1
Covt t; dt Covt yt; � yt;

dyt 2:27
0
where Covt ; denotes the covariance conditional on information available

in period t.
With a linear model, the changes dt and dyt caused by the change dijt
are deterministic and independent of t and yt , so that the covariance
terms in (2.27) vanish. The change dXt in the random variable Xt is
identical to the change dXtjt in the conditional forecast Xtjt . Then the
optimality criterion (2.22) applies, and it is suf®cient to think in terms of
conditional forecasts. With a nonlinear model, the changes dt and dyt are
stochastic and depend on t and yt and therefore the covariance terms in
(2.27) are not necessarily zero. In order to see this, note that with the linear
model (2.19), the change dXt from a change ditsjt 0 s < � 1) is given by
dXt dXtjt
dXt ditsjt ditsjt A�1�s B ditsjt
dits ditsjt
where we use that dXt =dits dXtjt =ditsjt A�1�s B, so dXt is determi-
nistic and independent of Xt (since the policy multipliers are constant).
With the non-linear model 2.26, the same change is
!
Y @MXtsr; itsr; utsr1 @MXts; its; uts1
r�s�1
dXt ditsjt
r1
@X @i
so that dXt is stochastic and generally correlated with Xt (since the policy
multipliers dXt =ditsjt are now endogenous and not constant).35
With a non-linear model, the optimal reaction function is nonlinear,
it f Xt
and it will generally depend on the covariances uu . The covariance terms in
(2.26) also imply that the optimal policy may imply a bias, in the sense that it
is optimal to, on average, either overshoot or undershoot the in¯ation target
(also, the optimal average output gap may not be zero).
Doing forecast targeting under a non-linear model and selecting the
instrument such that (2.17) is minimised would still imply a non-linear
reaction function, since the conditional forecasts would be non-linear
functions of the instrument path. It would imply disregarding the optimal
bias, though. How costly forecast targeting would be relative to the optimal
policy would, of course, depend on the degree of non-linearity in the
transmission mechanism, which is an empirical question. My reading of the
literature is that there is considerable controversy about the extent and
the relevance of any nonlinearity (see for instance Gordon, 1997, and Isard
and Laxton, 1996).
Non-additive uncertainty
So far, only additive uncertainty has been considered, appearing as the
additive shock ut1 in (2.8). Uncertainty about parameters in the model ± that
is, uncertainty in the coef®cients of the matrices A and B in (2.8) ± results in
multiplicative uncertainty, an example of non-additive uncertainty. Multi-
plicative uncertainty has consequences similar to non-linearity, in that
certainty-equivalence no longer holds, even if the model remains linear and
the loss function is quadratic.
Assume that the model is linear, that there are no forward-looking variables
and that the predetermined variables are observable, but assume now that the
model has time-varying stochastic parameters, with known stochastic
properties. Then the model can be written
Xt1 At1 Xt Bt1 it ut1 ; 2:28
where At and Bt are stochastic processes with known stochastic properties. It

follows that Xt can be written
! !
Y
�1 X
�1 Y
�s�1 X
Y
�s
Xt At1s Xt At1sr Bts1 its Atsr uts
s0 s0 r1 s1 r1
Then the policy multipliers,

!
dXt Y
�s�1
At1sr Bts1
dits r1
for 0 s � 1, are stochastic. It follows that a change in the interest rate

path, dijt , results in stochastic changes dt and dyt in in¯ation and output.
Once more, the covariance terms in (2.27) do not vanish, and certainty-
equivalence no longer applies. As shown in the classic paper by Brainard
(1967) and, for instance, in a more recent application to strict in¯ation
targeting in Svensson (1999c), the optimal instrument response to disturb-
ances usually becomes more cautious, and a bias enters such that the optimal
policy implies that the in¯ation either overshoots or undershoots the in¯ation
target. However, some covariance patterns for the parameters, as well as
uncertainty about the in¯ation persistency, may make the optimal instrument
response more sensitive to disturbances than in the absence of parameter
uncertainty (see So È derstro
È m, 1999a).
Monetary policy under model uncertainty is currently a very active research
area, with a number of recent papers.36 Practically all work is directed towards
the understanding of how the optimal reaction function changes with model
uncertainty. But how should practical monetary policy incorporate model
uncertainty? This is the issue to be discussed next.
Generalised forecast targeting: distribution forecast targeting instead of mean

forecast targeting
One possibility for handling non-linearity and model uncertainty might seem
to be to commit, once and for all, to a simple instrument rule, either to an
optimal reaction function that minimises the expected intertemporal loss,
taking Bayesian priors on the nature of the model uncertainty into account, or
a simple rule with reasonable robustness properties across potential models.
However, I believe the objections to this solution raised on p. 71±73 apply
with even greater force under non-linearity and model uncertainty, since the
unavoidability and desirability of judgemental adjustments, extra-model
information and updated Bayesian priors are even more apparent. Therefore,
I believe that the solution must be found elsewhere, namely in a generalisation
of forecast targeting.
Non-linearity and model uncertainty both imply that certainty-equivalence
does not apply. This, in turn, means that it is not optimal to rely on forecast
targeting with conditional mean forecasts ± that is, using conditional mean
forecasts as intermediate variables. Thus mean forecast targeting is not
optimal. Still, we can consider a kind of generalised forecast targeting,

distribution forecast targeting, as a way of handling non-linearity and
multiplicative uncertainty.
Distribution forecast targeting simply consists of constructing conditional
probability distributions of the target variables instead of means only. For any
random variable t , let t denote the random path t ; t1 ; ::: (recall that jt
denotes the path of the conditional means tjt ; t1jt ; :::). Then, for a given
interest rate path ijt , the central bank staff should construct the joint
conditional density function of the random path of in¯ation and the output
gap, denoted 't t ; yt � y t ; ijt , conditional upon all information available in
period t and the interest rate path ijt .
Then, this conditional probability distribution is used to evaluate the loss
function (2.1) with (2.7). This can either be done numerically, or informally by
the decision making body of the bank being presented with the probability
distributions for a few alternative interest rate paths and then deciding which
path and distribution provides the best compromise.
This alternative is actually much more feasible than many readers might
think. The Bank of England and the Sveriges Riksbank have already developed
methods for constructing con®dence intervals for the forecasts published in
their In¯ation Reports (see Blix and Sellin, 1998 and Britton, Fisher and
Whitley, 1998). The Bank of England presents fan charts for both in¯ation and
output, and the Sveriges Riksbank gives con®dence intervals for its in¯ation
forecasts.37 Furthermore, scrutiny of the motivations for interest rate changes
(including the minutes for the Bank of England's Monetary Policy Committee
and the Riksbank's Executive Board) indicate that both banks occasionally
depart from certainty-equivalence and take properties of the whole distribu-
tion into account in their decisions, for instance, when the risk is unbalanced
and `downside risk' differs from ùpside risk'.
The result of distribution forecast targeting can be compared with what
would result if only conditional mean forecasts were considered. In practice,
there may not be a big difference in the resulting instrument setting, in which
case there is not much point in letting distribution forecast targeting replace
mean forecast targeting. Distribution forecast targeting seems meaningful and
worth the effort only if it results in signi®cantly different, and more adequate,
instrument settings than mean forecast targeting. This remains to be
examined.
Intermediate targeting
When would intermediate targeting be optimal? Assume, for simplicity, strict
in¯ation targeting, (2.4). Consider, for simplicity, the model without forward-
looking variables, (2.19). Decompose the vector of predetermined variables
according to Xt t ; X02t ; X03t 0 , where the ®rst element is in¯ation and the rest
of the variables are decomposed into two vectors, X2t and X3t . Suppose that
the two vectors X2t and X3t can be chosen such that the model (2.19) of the
transmission mechanism ful®ls
2 3 2 32 3 2 3
t1 0 A12 0 0
4 X2;t1 5 4 A21 A21 A23 54 X2t 5 4 B2 5it ut1
X3;t1 A31 A32 A33 X3t B3
that is, where the A and B matrices are such that A11 0, A13 0 and B1 0.
Then in¯ation ful®lls
t1 A12 X2t u1;t1; 2:29
and is exclusively determined by variables X2t ; and variables X2t are the only
predictors of in¯ation (aside from the zero-mean exogenous shock u1;t1 ).
Under these assumptions, the instrument it affects in¯ation exclusively by ®rst
affecting X2;t1 and then by X2;t1 affecting t2 . Schematically, we have
it ! X2;t1 ! t2
Because of this property of X2t ; its elements can be called intermediate variables.
Let X2 ful®ll
A12 X2 : 2:30
Substituting (2.29) and (2.30) into (2.4) for t 1 and taking the expectation in
period t result in
Lt1jt 12t1jt � 2 21u1

2
X2t � X2 0 WX2t � X2 12u1
2
2
where u1 is the variance of u1t and the weight matrix W ful®ls W 12 A012 A12 .
Clearly, using the period loss function
L~ t X2t � X2 0 WX2t � X2
is equivalent to using (2.4). Now we can call X2t intermediate target variables, X2
intermediate target levels, and minimising (2.1) with L~ t instead of (2.7) we can
call intermediate targeting. We thus have a situation where intermediate
targeting is as good as strict in¯ation targeting.
In particular, assume that in¯ation is exclusively determined by money
growth according to
t1 mt u1;t1
where mt mt � mt�1 and mt is the log of a monetary aggregate. Then, the
instrument exclusively affects in¯ation via ®rst affecting money-growth ± that
is,
it ! mt1 ! t2
Thus, let X2t mt , A12 1, X2 and W 12 ; and strict in¯ation
targeting can be replaced by strict money growth targeting with the period
loss function
1
L~ t mt � 2
2
Both kinds of targeting will be equivalent.
In the example above, the transmission mechanism is recursive in a special
way, such that the target variables (in the above case only in¯ation) are
determined only by a set of intermediate variables (the only exception being
zero-mean exogenous shocks). Clearly, this is an extremely special case. In the
real world, and in reasonable models, the transmission mechanism is too
complex for intermediate variables in this sense to exist ± that is, the
transmission mechanism is not recursive in the above sense.38
Therefore, intermediate targeting in general, and monetary targeting in
particular, is not a good monetary policy strategy. However, there is one
exception to the general non-existence of intermediate variables. As discussed
in Svensson (1997a, 1999b), one set of intermediate variables always exists,
namely conditional forecasts. For any vector Yt of target variables, we can write
Yt Ytjt "t
where Ytjt is a conditional forecast of Yt , conditional on information

available in period t, and "t1 is an error term uncorrelated with information in
period t. Formally, conditional forecasts can be seen as intermediate variables,
and forecast targeting can be seen as intermediate targeting. As King (1994)
stated early in the history of in¯ation targeting, in¯ation targeting means
having in¯ation forecasts as intermediate targets.
4 Lessons for the Eurosystem
De®ning price stability

In its ®rst announcement of its monetary policy strategy, on 13 October 1998,
the Eurosystem (European Central Bank, 1998a) stated:
Price stability shall be de®ned as a year-on-year increase in the Harmonised

Index of Consumer Prices (HICP) for the euro area of below 2%.
Thus, only an upper bound of 2 per cent was unambiguously announced.

About a month later, the Eurosystem clari®ed that ìncrease' in this de®nition
shall be interpreted as excluding de¯ation. This might seem to imply that the
lower bound for in¯ation is zero. On the other hand, the Eurosystem also
stated that it had not announced a lower bound, giving the reason that the size
of any measurement bias in the HICP is not known. On 1 December, 1998, the
Eurosystem (European Central Bank, 1998b) announced a monetary reference

value for M3 of 4.5 per cent. Then, subtraction of the sum of its estimates of
potential output growth and trend decline in velocity from 4.5 per cent
revealed that it had applied an in¯ation target of 1.5 per cent. If this is the
middle of an interval with an upper bound of 2 per cent, that interval is
obviously 1±2 per cent. Hence, the Eurosystem seems to have implicitly
announced a lower bound of 1 per cent.
Clearly, observers should not have to piece together the de®nition from
different statements, including the announcement of the reference value. Any
remaining ambiguity with regard to the de®nition seems to serve no purpose
and should be eliminated. Instead of the ambiguous and asymmetric
statement `below 2%', the Eurosystem had better state the unambiguous
and symmetric 1.5 per cent as its in¯ation target, or the interval 1±2 per cent,
possibly with the addendum that this de®nition may be somewhat modi®ed
when more evidence about the quality of the HICP becomes available. Absent
such a statement, it seems that observers should currently interpret the
Eurosystem as having an in¯ation target of 1.5 per cent, and evaluate it
accordingly.
As argued on pp. 62±5, it is by no means clear that interpreting price stability
as an in¯ation target should be the ®nal word, since that implies accepting a
unit root and non-stationarity of the price level, making `price stability' a
misnomer. After some ®ve or ten years of successful in¯ation targeting, I
believe the Eurosystem should seriously consider the pros and cons of moving
to price-level targeting.
Maintaining price stability

In addition to the de®nition of price stability, on 13 October, the Eurosystem
announced what was later called the `two pillars' of its strategy:
. à prominent role for money with a reference value for the growth of a
monetary aggregate' and
. à major role for a broadly-based assessment of the outlook for future price
developments'.39
With regard to the role of money, the Eurosystem has emphasised that the
reference value should not be interpreted as an intermediate target for money
growth. Indeed, it has rejected monetary targeting, on the grounds that the
relationship between money and prices may not be suf®ciently stable, and that
it is not clear that the monetary aggregates with the most stable relationship is
suf®ciently controllable in the short run. As Issing (1998) summarises:
In these circumstances, relying on a pure monetary targeting strategy

would constitute an unrealistic, and therefore misguided, commitment.
Instead, the Eurosystem plans to use money growth as an indicator of `risks

to price stability,' such that deviations of current money growth from the
reference value signals risks to price stability.
The reference value will be derived in a manner that ensures, as far as

possible, that deviations of monetary growth from the value will signal risks
to price stability. In the ®rst instance, such a deviation will prompt further
analysis to identify and interpret the economic disturbance that caused the
deviation, and evaluate whether the disturbance requires a policy move to
counter risks to price stability. (Issing, 1998)
As I argue in some detail in Svensson (1999d), there is little ground for such a
prominent role for money. It is easily shown in the simple and conventional
model used there, that such a money growth indicator will be a relatively useless
indicator of risks to price stability ± and, indeed, mostly a noisy indicator of the
deviation of current in¯ation from the in¯ation target. As argued on p. 79±82, the
weight on money as an indicator should be strictly determined by its predictive
power in forecasting in¯ation. As argued on p. 81±82 and demonstrated in some
detail in Svensson (1999d), the best indicator of `risks to price stability' is an
in¯ation forecast conditional on an unchanged interest rate.40
I believe it worthwhile to look more closely at the Eurosystem's arguments
in favour of a prominent role to money, stated very clearly in Issing (1998).
Three arguments for giving a prominent role for money are provided: (1)
Ìn¯ation is fundamentally monetary in origin over the longer term', (2) Ìt
creates a ®rm `nominal' anchor for monetary policy and therefore helps to
stabilise private in¯ation expectations at longer horizons', and (3) `[It]
emphasises the responsibility of the ESCB for the monetary impulses to
in¯ation, which a central bank can control more readily than in¯ation itself'.
With regard to argument (1), it is based on the empirical high long-run
correlation between money and prices. This correlation, however, holds in any
model where demand for money is demand for real money, for instance in the
simple model used in Svensson (1999d) to demonstrate the inferiority of the
Eurosystem's money growth indicator. The correlation is actually a relation
between two endogenous variables, and says nothing about causality. The
direction of causality is determined by the monetary policy pursued. Under
strict monetary targeting, when the central bank aims at maintaining a given
money growth rate regardless of what happens to prices, money growth
becomes exogenous in the relation and causes in¯ation, which is endogenous.
Under in¯ation targeting, when the central bank aims at maintaining a given
in¯ation rate regardless of what happens to money, in¯ation becomes
exogenous in the relation and causes money growth, which is endogenous.
Hence, argument (1) is neutral to the monetary strategy.
With regard to argument (2), it seems that the de®nition of price stability
provides the best nominal anchor and is the best stabiliser of in¯ation
expectations. Emphasising a second nominal anchor seems redundant and

even misguided, since more than one nominal anchor could confuse, rather
than stabilise, private expectations.
With regard to argument (3), it is not clear how `monetary impulses to
in¯ation' can be de®ned in any unambiguous and useful way. I am not sure
the argument means anything but monetary aggregates being easier to control
than in¯ation. Furthermore, it seems obvious that the Maastricht Treaty does
not assign `stability of monetary impulses to in¯ation' as the primary objective
of the Eurosystem; it assigns `price stability', period.
Thus, the ®rst pillar is unlikely to provide much support for the maintenance
of price stability. Instead, the maintenance of price stability must rely on the
second pillar, the `broadly-based assessment of the outlook for future price
developments'. This is, of course, nothing but a long euphemism for in¯ation-
forecast targeting. Indeed, I believe the success of the Eurosystem maintenance
of price stability depends strongly on its learning to do forecast targeting, as
practised by an increasing number of in¯ation-targeting central banks.
Strangely enough, Eurosystem statements argue that in¯ation-forecast
targeting would be unsuitable for the Eurosystem, on the grounds that
forecasting in¯ation will be dif®cult and that the understanding of the
transmission mechanism is imperfect. To quote Issing (1998) further:41
In the uncertain environment likely to exist at the outset of Monetary

Union, forecasting in¯ation will be dif®cult, not least because of the many
conceptual, empirical and practical uncertainties faced by the ESCB at the
start of Stage Three. Forecasting models estimated using historic data may
not offer a reliable guide to the behaviour of the euro area economy under
Monetary Union. Forecast uncertainty is likely to be relatively large.
Forecasting in¯ation requires thorough knowledge of the properties of
the new euro area-wide data series and experience and understanding of the
transmission mechanism of monetary policy in the new euro area
economy. Both are likely to be quite different from what we have been
used to in the existing environment of eleven distinct national economies
prior to Monetary Union.
Certainly, forecasting and forecast targeting will not be easy, and forecast
uncertainty is likely to be relatively large. Nevertheless, forecasting is simply
necessary, given `the need for monetary policy to have a forward-looking,
medium-term orientation' that Issing and the Eurosystem emphasise.
Furthermore, forecast targeting implies using existing information in the
most ef®cient and ¯exible way. It incorporates both model and extra-model
information, allows judgemental adjustments, takes additive uncertainty for
granted and even allows imperfect understanding of the transmission
mechanism and model uncertainty, as I have tried to explain in this chapter.
Of course, the less the uncertainty and the better the understanding of the
transmission mechanism, the more successful forecast targeting is likely to be.

But this does not mean that there is some better way of maintaining price
stability if there is more uncertainty and less understanding of the
transmission mechanism.42
5 Conclusions
This chapter argues that forecast targeting is the best way of maintaining price
stability, on the grounds that with lags and uncertainty in the transmission
mechanism, forecast targeting is the most ef®cient and ¯exible way of using
available information. By generalising forecast targeting from mean forecast
targeting to distribution forecast targeting, it should also be the best way of
handling model uncertainty. Indeed, I believe the current best practice in
central banks' maintaining price stability must be understood as forecast
targeting.
The chapter has, so far, discussed only the framework for policy decisions
and not at all the central bank's communication, degree of transparency and
degree of accountability. Under forecast targeting, the conditional forecasts for
in¯ation and the output gap are the crucial inputs in the policy decision.
Therefore, policy decisions are best explained and motivated, and policy is best
understood and anticipated by the public, with reference to these conditional
forecasts. This has the bene®cial effect that any criticism of the policy must be
more speci®c: for instance, is it the target or the central bank's forecast that is
wrong? Furthermore, making these forecasts public provides the best
opportunity for outside observers to monitor and evaluate the central bank's
policy, and making sure that its decisions are consistent with its objective.
Then policy can be evaluated almost in real time, without waiting some two
years to see the outcome of an in¯ation rate that is, by then, contaminated by
a number of intervening shocks. Finally, making the forecasts public provides
the strongest incentives for the central bank to improve its competence and do
the best possible job.
These are strong arguments in favour of making these forecasts public, a
practice already followed by the Reserve Bank of New Zealand, the Bank of
England, and the Sveriges Riksbank. Against this background, the Eurosys-
tem's refusal to publish its forecasts, citing far from convincing arguments,43 is
very dif®cult to understand, except perhaps as an expression of an initial lack
of con®dence and experience and a desire to further improve its competence
before going public (but if so, why not announce that the forecasts will
eventually be made public?). I see no reason why the Eurosystem should not
aim for the current best standard of transparency, as demonstrated by the
central banks already mentioned.
Notes
1. The paper which this chapter is based upon was presented at the Bundesbank
conference on `The Monetary Transmission Process: Recent Developments and
Lessons for Europe', 26±27 March, 1999. I thank the discussants Mervyn King and
JoseÂ Vi~nals, and Claes Berg, Donald Brash, John Faust, Torsten Persson, Anders
Vredin and participants in seminars at IIES and Sveriges Riksbank for comments.
Special thanks are due to Jon Faust and Dale Henderson. Over the years, I have
bene®ted a great deal from many discussions with them on monetary policy and
from their very constructive (sometimes relentless) criticism of previous work of
mine. Dale has also directed me to an early, very relevant, literature. I also thank
Christina Lo È nnblad for editorial and secretarial assistance and Marcus Salomonsson
for research assistance. Needless to say, the views expressed and any errors are my
own responsibility.
2. This literature includes Bernanke et al (1998), Bernanke and Mihov (1997), Clarida,
Gali and Gertler (1998a), Clarida and Gertler (1997), Laubach and Posen (1997),
Neumann (1997), von Hagen (1995) (note a crucial typo: the coef®cient for money
supply in Table 1 should be 0.07 instead of 0.7).
3. `Strict' and `¯exible' targeting is de®ned below.
4. If arguments in favour of a small positive in¯ation rate is accepted, an upward-
sloping price-level target path may be preferable.
5. An interesting issue is to what extent the degree of nominal rigidity depends on
whether there is in¯ation or price-level targeting.
6. This result requires at least moderate output persistence with a Lucas-type Phillips
curve, and does not hold for a Lucas-type Phillips curve without persistence. Kiley
(1998) shows that the result does not hold for a Calvo-type Phillips curve without
persistence.
7. See, for instance, McCallum and Nelson (1999) and Williams (1997).
8. As in¯ation-targeting central banks, like other central banks, also seem to smooth
instruments, the loss function (2.7.) may also includes the term it � it�1 2 with
> 0.
9. See box on p. 26 in Sveriges Riksbank (1997) as well as Heikensten and Vredin
(1998).
10. àctual in¯ation will on occasions depart from its target as a result of shocks and
disturbances. Attempts to keep in¯ation at the in¯ation target in these
circumstances may cause undesirable volatility in output'.
11. See Bank of England (1998), para. 40: `[I]n any given circumstances, a variety of
different interest rate paths could in principle achieve the in¯ation target. What
factors were relevant to the preferred pro®le of rates?... There was a broad consensus
that the Committee should in principle be concerned about deviations of the level
of output from capacity'.
12. The Reserve Bank's target was previously de®ned in terms of a somewhat complex
underlying in¯ation rate. In the Policy Target Agreement of December 1997, there
was a change to the more transparent CPIX.
13. On the other hand, the argument that in¯ation increases capital market
distortions, examined in Feldstein (1997, 1999), would, under the assumption of
unchanged nominal taxation of capital, motivate a zero or even a negative in¯ation
target.
14. For reasons explained in Gordon (1996), I believe that Akerlof, Dickens and Perry
(1996) reach too pessimistic a conclusion. On the other hand, their data is from the
United States and Canada, and downward nominal wage rigidity may be more
relevant in Europe. The conclusions of Orphanides and Wieland (1998) are
sensitive to assumptions about the size of shocks and the average real interest rate;
the latter is taken to be 1 per cent for the United States. If the average real rate is
higher in Europe, and the shocks not much larger than in the United States, non-
negative interest rates may be of less consequence in Europe. Wolman (1998a,
1998b) provides a rigorous examination of the consequences of non-negative
interest rates in a more explicit model, and ®nds relatively small effects.
15. Also, Wolman (1998b) ®nds that a reaction function responding to price-level
deviations from a price-level target (rather than in¯ation deviation from an
in¯ation target) has good properties for low in¯ation rates.
16. Note that this simple ®rst-order condition only arises if the variables in Xt are
predetermined.
17. There is an additional philosophical objection to once-and-for-all commitment:
how come the once-and-for-all commitment can be done in period 0? Why was it
not already done before, so nothing remains to be committed to in period 0? Why
is there something special about period 0?
18. I found this appropriate quote in Budd (1998).
19. See Budd (1998) for an interesting and detailed discussion of the advantages of
explicitly considering forecasts rather than formulating reaction functions from
observed variables to the instrument.
20. It is important that these expectations are conditional on the central bank's model,
and hence are `structural', rather than being private sector expectations, in order to
avoid the problems of non-existence or indeterminacy of equilibria, arising from
responding mechanically to private sector expectations, as has been emphasised in
Woodford (1994) and further discussed in Bernanke and Woodford (1997).
21. The conditional forecasts for arbitrary interest-rate path derived in Svensson
(1998a, Appendix A] assume that the interest rate paths are `credible' that is,
anticipated and allowed to in¯uence the forward-looking variables. Leeper and Zha
(1999) discuss an alternative way of constructing forecasts for arbitrary interest rate
paths, by assuming that these interest rate paths result from unanticipated
deviations from a normal reaction function.
22. The consequences of imposing the restriction of time-consistency of it remain to be
examined. That is, that the elements itjt in it shall be consistent with the decision
in period t conditional on Xtjt (see n. 6, p. 93).
23. Note that one way of taking the discretionary nature of decision making into
account is to set it in period t under the restriction that the reaction function that
will apply in period t for 1 will be itjt ftjt Xtjt , where ftjt is the
reaction function that is likely to result from the decision in period t .
24. Note that an equation like it gXtTjt , where XtTjt is a model-consistent forecast
(including this equation), especially in a model with forward-looking models, is a
rather complex equilibrium condition. For reasons detailed in Svensson (1999b,
section 2.3.1), I am rather sceptical about these equilibrium conditions as reaction
functions representing in¯ation targeting.
25. See Tinsley (1975) and Reifschneider, Stockton and Wilcox (1997) for further
discussion of judgemental adjustments.
26. Mervyn King has emphasised that it is important that the decision making body of
the central bank agrees with the forecast. This requires iterations between the staff
and the decision making body, with the decision making body having the last say
on the forecast.
27. Thus, for the `strict' case with in¯ation as the only argument in the loss function,
the four loss functions are (1) Lt 12 t � 2 , (2) Lt jt � j, (3)
Lt k � t � , where x is the so-called R 1Dirac delta function with the
properties x 0 for x 6 0; 0 1; and �1 xdx 1, and (4) Lt 0 for
jt � j a > 0, Lt k > 0 for jt � j > a.
28. Another problem with reporting the mode forecast is evident in the hypothetical
case when the forecast is bimodel with approximately equal probability density at
the two modes.
29. Two separate arguments are sometimes presented in favour of emphasising the
mode forecast. First, in presenting and discussing the forecast, it may often be natural
and intuitive to consider a most likely scenario together with one or two alternative
scenarios. The most likely scenario would then correspond to the mode forecast.
Second, before that stage, in constructing the forecast, it may be practical to start
with a most likely scenario and then add various uncertainties and complications
later on. Whereas the ®rst argument may be a legitimate argument in favor of the
mode, the second is not, since the presentation and the construction can be
independent.
Furthermore, the mode and the median have the property that they are not affected
by outlines, which may or may not be an advantage, depending on one's view.
30. See Orphanides (1998) and Smets (1998) for recent related work.
31. In a steady state, the Kalman gain is given by K PD0 DPD0 vv �1 , where the
covariance matrix P of the forecast errors Xt � Xtjt�1 is given by
P MP � PD0 DPD0 vv �1 DPM 0 uu , where M is the transition matrix in the
transition equation Xt1 MXt ut1 :
32. That such a money growth indicator is unsuitable on its own is fairly obvious, since
money is not the only, not even the major, predictor of in¯ation at the horizons
relevant for monetary policy (see Estrella and Mishkin, 1998). What is perhaps less
obvious is that the money growth indicator is unsuitable even for a completely
stable money-demand function without velocity shocks (see Svensson, 1999d, and
Rudebusch and Svensson, 1999b).
33. More speci®cally, with forward-looking variables, the interest rate is kept
unchanged for a few periods (four±six quarters, say), but then, there is a shift to
`normal' policy, or to some policy stabilizing in¯ation and determining the future
forward-looking variables.
34. A non-quadratic loss function would also imply that certainty-equivalence no
longer holds (see p. 78±79).
35. Note that we use the convention that tr s r 1 for t < s:
36. Recent work on and discussion of monetary policy under model uncertainty
includes Blinder (1998), Cecchetti (1997), Clarida, Gali and Gertler (1998b), Estrella
and Mishkin (1998), Levin, Wieland and Williams (1999), McCallum (1997),
Onatski and Stock (1998), Peersman and Smets (1998), Rudebusch (1998), Sack
(1988), Sargent (1998) Smets (1998), So È derstro
È m (1999a, 1999b), Stock (1998),

Svensson (1999c), and Wieland (1996, 1998).
In contrast to the standard Brainard result in favour of caution, work on so-called

robust control (where the reaction function is chosen so as to minimise expected
loss for the most unfavourable model) indicates that model uncertainty may well
result in more aggressive optimal responses (see, for instance, Sargent, 1998, and
Stock, 1998).
37. The Bank of England's fan charts for in¯ation and output should probably be
interpreted as marginal distributions. However, since the distributions for in¯ation
and the output gap are unlikely to be independent, distribution forecast targeting
requires the joint distribution to be conveyed. This may require some further
innovation in display, beyond the already beautiful fan chart.
38. As discussed in Wallis (1999), the Bank of England's fan charts present prediction
intervals that differ from normal con®dence intervals, central prediction
intervals. The Sveriges Riksbank, however, presents normal con®dence intervals,
(see Blix and Sellin, 1998). Both banks present the mode as their point forecast,
whereas it seems to me that it would be more natural and consistent with the
theory to present the mean (or, in distribution forecast targeting, at least the
median).
39. See Bryant (1980), Friedman (1975), Kalchbrenner and Tinsley (1975) and Kareken,
Muench and Wallace (1973) for this and other arguments against intermediate
targeting in general and monetary targeting in particular. See Rudebusch and
Svensson (1999b) for simulations of monetary targeting in the U.S. with lessons for
the Eurosystem. These simulations show that monetary targeting in the United
States would be quite inef®cient compared to ¯exible in¯ation targeting, in the
sense of bringing higher variability of both in¯ation and the output gap.
40. Since the ®rst version of this paper was written, an extensive discussion and
motivation of Eurosystem strategy has been presented by Angeloni, Gaspar and
Tristani (1999).
41. Rudebusch and Svensson (1999b) examine monetary targeting in an empirical
model of in¯ation, output and money for US data and draw some lessons for the
Eurosystem. They ®nd that monetary targeting would be very inef®cient compared
to in¯ation targeting, in the sense of increasing the variability of both in¯ation and
output. Counter to conventional wisdom, this is the case also if money demand
shocks are set to zero so the money demand is completely stable.
Gerlach and Svensson (1999) examine the indicator properties of monetary
aggregates for the euro area. Somewhat surprisingly, they ®nd considerable
empirical support for the so-called P model of Hallman, Porter and Small
(1991), adapted to Germany by To È dter and Reimers (1994). This implies that
monetary aggregates, in the form of the `real money gap', the gap between current
real balances and long-run equilibrium real balances, has considerable predictive
power for future in¯ation. They ®nd little or no empirical support for the
Eurosystem's money growth indicator, though.
Indeed, the theoretical analysis in Svensson (2000a) shows that the P model,
although emphasising the role of the real money gap in forecasting and controlling
in¯ation, does not provide any support for a Bundesbank-style money growth
target or a Eurosystem-style money-growth indicator.
42. See also Angeloni, Gaspar and Tristani (1999).
43. As I argue in Svensson (1999d), it may be sobering to recall that the introduction of
in¯ation targeting in the United Kingdom, Sweden and Finland occurred under
rather dramatic circumstances. The countries went through dramatic boom±bust
experiences, very serious banking and ®nancial sector crises, and a dramatic sudden
shift from a ®xed exchange rate to a new monetary policy regime with a ¯oating
exchange rate. Furthermore, this occurred in a situation with very low credibility
for monetary policy, with high and unstable in¯ation expectations, much above
the announced in¯ation targets. At least for Sweden (where I am naturally more
informed) the central bank's commitment to the ®xed exchange rate was so strong,
that there was no contingency planning. When the krona was ¯oated in November
1992, the new in¯ation-targeting regime, which was announced in January 1993,
had to be conceived from scratch (although, of course, with the bene®t of the
experiences mainly from New Zealand and Canada). It is not easy to rank
dif®culties and uncertainty about the transmission mechanism, but it seems to me
that the dif®culties facing the Eurosystem are still not of the same magnitude as the
dif®culties that the central banks of the United Kingdom, Sweden and Finland were
facing. Since those central banks have, nevertheless, managed quite well, the odds
for the Eurosystem may be quite good, provided it adopts a similar framework for
policy decisions.
44. See, for instance, Duisenberg (1998): `publishing an in¯ation forecast would
obscure rather than clarify what the Governing Council is actually doing. The
public would be presented with a single number intended to summarise a thorough
and comprehensive analysis of a wide range of indicator variables. However, such a
summary would inevitably be simplistic. Moreover, because publishing a single
in¯ation forecast would be likely to suggest that monetary policy reacts
mechanistically to this forecast, publication might mislead the public and therefore
run counter to the principle of clarity'.
References
Akerlof, G. A., W. T. Dickens and G. L. Perry (1996), `The Macroeconomics of Low
In¯ation', Brookings Papers on Economic Activity, 1, 1±76.
Angeloni, I., V. Gaspar and O. Tristani (1999) `The Monetary Policy Strategy of the ECB',
ECB Working Paper.
Aoki, K. (1998). Òn the Optimal Monetary Policy Response to Noisy Indicators',
Princeton University, Working Paper.
Backus, D. and J. Drif®ll (1986) `The Consistency of Optimal Policy in Stochastic
Rational Expectations Models', CEPR Discussion Paper, 124.
Ball, L. (1997) Èf®cient Rules for Monetary Policy', NBER Working Paper, 5952.
ÐÐÐÐ (1999) `Policy Rules for Open Economies', in Taylor (1999).
Bank of Canada (1994) Economic Behavior and Policy Choice under Price Stability, Ottawa,
Bank of Canada.
Bank of England (1998) Minutes of the Monetary Policy Committee meeting on 4±5 March
1998, Bank of England.
Batini, N. and A.G. Haldane (1999) `Forward-looking Rules for Monetary Policy', in
Taylor (1999).
Berg, C. and L. Jonung (1999) `Pioneering Price Level Targeting: The Swedish Experience
1931±1937', Journal of Monetary Economics, 43, 3, 525±51.
Bernanke, B.S., T. Laubach, F.S. Mishkin and A.S. Posen (1998) In¯ation Targeting: Lessons
from the International Experience, Princeton University Press.
Bernanke, B.S., and I. Mihov (1997) `What Does the Bundesbank Target?', European
Economic Review, 41, 1025±54.
Bernanke, B.S., and F.S. Mishkin (1997) Ìn¯ation Targeting: A New Framework for
Monetary Policy?', Journal of Economic Perspectives; 11, 97±116.
Bernanke, B.S. and M. Woodford (1997) Ìn¯ation Forecasts and Monetary Policy',
Journal of Money, Credit, and Banking, 29, 653±84.
Blinder, A.S. (1998) Central Banking in Theory and Practice, Cambridge, MA, MIT Press.
Blix, M. and P. Sellin (1998) Ùncertainty Bands for In¯ation Forecasts', Sveriges Riksbank
Working Paper Series, 65.
Brainard, W. (1967) Ùncertainty and the Effectiveness of Policy', American Economic
Review, Papers and Proceeding, 57, 411±25.
Britton, E., P. Fisher and J. Whitley (1998) `The In¯ation Report Projections: Under-
standing the Fan Chart', Bank of England Quarterly Bulletin, 38, 30±7.
Bryant, R.C. (1980) Money and Monetary Policy in Interdependent Nations, Washington
D.C, Brookings Institution.
Bryant, R.C., P. Hooper and C.L. Mann (eds) (1993) Evaluating Policy Regimes: New
Research in Empirical Macroeconomics, Washington, DC, The Brookings Institution.
Budd, A. (1998) Èconomic Policy, with and without Forecasts', Bank of England Quarterly
Bulletin, 86, 379±84.
Cecchetti, S.G. (1995) Ìn¯ation Indicators and In¯ation Policy', NBER Macroeconomics
Annual, 10, 189±219.
ÐÐÐÐ (1997) `Central Bank Policy Rules: Conceptual Issues and Practical Considera-
tions', Federal Reserve Bank of New York, Working Paper.
Chow, G.C. (1975) Analysis and Control of Dynamic Economic Systems, New York, John
Wiley.
Clarida, R., J. Gali and M. Gertler (1998a) `Monetary Policy Rules in Practice: Some
ÐÐÐÐ (1998b) `The Science of Monetary Policy, A New Keynesian Approach', NBER
Working Paper 7147.
Clarida, R. and M. Gertler (1997) `How the Bundesbank Conducts Monetary Policy', in
C.D. Romer and D.H. Romer (eds), Reducing In¯ation: Motivation and Strategy, Chicago,
Chicago University Press.
Clark, P., D. Laxton and D. Rose (1996) Àsymmetry in the U.S. Output±In¯ation Nexus',
IMF Staff Papers, 43, 216±51.
Currie, D. and P. Levine (1992) `Should Rules Be Simple?', Economics and Planning, 25,
113±38, reprinted in Currie and Levine (1993).
ÐÐÐÐ (1993) Rules, Reputation and Macroeconomic Policy Coordination, Cambridge,
Cambridge University Press.
Debelle, G. and D. Laxton (1997) Ìs the Phillips Curve Really a Curve? Some Evidence for
Canada, the United Kingdom, and the United States', IMF Staff Papers, 44, 249±82.
Duisenberg, W.F.(1998) `The ESCB's Stability-oriented Monetary Policy Strategy', speech
in Dublin, 10 November < http://www.ecb.int >.
Estrella, A. and F.S. Mishkin (1997) Ìs there a Role for Monetary Aggregates in the
Conduct of Monetary Policy?', Journal of Monetary Economics, 40, 279±304.
ÐÐÐÐ (1998) `Rethinking the Role of NAIRU in Monetary Policy: Implications of
Model Formulation and Uncertainty', NBER Working Paper, 6518.
European Central Bank (1998a) À Stability-oriented Monetary Policy Strategy for the
ESCB', Press Release, 13 October < http://www.ecb.int >.
ÐÐÐÐ (1998b) `The Quantitative Reference Value for Monetary Growth', Press
Release, 1 December < http://www.ecb.int >.
ÐÐÐÐ (1999) `The Stability-oriented Monetary Policy Strategy of the Eurosystem',
ECB Monthly Bulletin, January, < http://www.ecb.int >, 39±50.
European Monetary Institute (1997) The Single Monetary Policy in Stage Three: Elements of
the Monetary Policy Strategy in the ESCB, European Monetary Institute, Frankfurt.
Federal Reserve Bank of Kansas City (1996) Achieving Price Stability, Kansas City, Federal
Reserve Bank of Kansas City.
Feldstein, M. (1997) `The Costs and Bene®ts of Going from Low In¯ation to Price
Stability', in C.D. Romer and D.H. Romer (eds), Reducing In¯ation: Motivation and
Strategy, Chicago, University of Chicago Press.
Feldstein, Martin (ed.) (1999) Costs and Bene®ts of Price Stability, Chicago, Chicago
University Press.
Fischer, S. (1994) `Modern Central Banking', in F. Capie, C. Goodhart, S. Fischer and

N. Schnadt, The Future of Central Banking, Cambridge, Cambridge University Press.
ÐÐÐÐ (1996) `Why Are Central Banks Pursuing Long-Run Price Stability?', in Federal
Reserve Bank of Kansas City (1996).
Fisher, I. (1934) Stable Money; published 1935 in Britain under the title of Stable Money: A
History of the Movement, London, Allen & Unwin.
Friedman, B.M. (1975) `Targets, Instruments, and Indicators of Monetary Policy', Journal
of Monetary Economics, 1, 443±73.
Gerlach, S. and G. Schnabel (1998) `The Taylor Rule and Average Interest Rates in the
EMU-11 Area: A Note', Bank for International Settlements, Working Paper.
Gerlach, S. and L.E.O. Svensson (1999) `Money and In¯ation in the Euro Area: A Case for
Monetary Indicators?', Working Paper.
Gordon, R.J. (1996) `Comments and Discussion', Brookings Papers on Economic Activity 1,
60±6.
ÐÐÐÐ (1997) `The Time-Varying NAIRU and its Implication for Economic Policy',
Journal of Economic Perspectives, 11(1), 11±32.
Greenspan, A. (1994) `Discussion', in F. Capie, C. Goodhart, S. Fischer and N. Schnadt,
The Future of Central Banking, Cambridge, Cambridge University Press.
Hall, R.E. (1984) `Monetary Strategy with an Elastic Price Standard', in Price Stability and
Public Policy, Kansas City, Federal Reserve Bank of Kansas City, 137±59.
Hallman, J.J., R.D. Porter and D.H. Small (1991) Ìs the Price Level Tied to the M2
Monetary Aggregate in the Long Run?', American Economic Review, 81, 841±58.
Heikensten, L. and A. Vredin (1998) Ìn¯ation Targeting and Swedish Monetary Policy ±
Experience and Problems', Sveriges Riksbank Quarterly Review, 4/1998, 5±33.
HM Treasury (1997) `Remit for the Monetary Policy Committee', News Release, 12 June.
Isard, P. and D. Laxton (1996) `Strategic Choice in Phillips Curve Speci®cation: What if
Bob Gordon is Wrong?', Paper prepared for a conference on European unemployment
in Florence, Italy.
Issing, O. (1998) `The European Central Bank at the Eve of EMU', speech in London,
26 November.
Kalchbrenner, J.H. and P.A. Tinsley (1975) Òn the Use of Optimal Control in the Design
of Monetary Policy', Federal Reserve Board, Special Studies Paper, 76.
Kareken, J.H., T. Muench and N. Wallace (1973) Òptimal Open Market Strategy: The Use
of Information Variables', American Economic Review, 63, 156±72.
Kiley, M.T. (1998) `Monetary Policy under Neoclassical and New-Keynesian Phillips
Curves, with an Application to Price Level and In¯ation Targeting', Federal Reserve
Board, Working Paper.
King, M.A. (1994) `Monetary Policy in the UK', Fiscal Studies, 15, (3), 109±28.
ÐÐÐÐ (1996) `How Should Central Banks Reduce In¯ation? ± Conceptual Issues', in
Federal Reserve Bank of Kansas City (1996).
ÐÐÐÐ (1997) `Changes in UK Monetary Policy: Rules and Discretion in Practice',
Journal of Monetary Economics, 39, 81±97.
Laubach, T. and A.S. Posen (1997) Disciplined Discretion: Monetary Targeting in Germany
and Switzerland, Essays in International Finance, 206, Princeton University.
Leeper, E.M., and T. Zha (1999) `Modest Policy Intervention', Indiana University,
Bloomington, Working Paper.
Leiderman, L. and L.E.O. Svensson (eds) (1995) In¯ation Targets, London, Centre for
Economic Policy Research.
Levin, A., A.V. Wieland and J.C. Williams (1999) `Robustness of Simple Monetary Policy
Rules under Model Uncertainty', in Taylor (1999).
LeRoy, S.F. and R.N. Waud (1977) Àpplications of the Kalman Filter in Short-run
Monetary Control', International Economic Review, 18, 195±207.
Lowe, P. (ed.) (1997) Monetary Policy and In¯ation Targeting, Sydney: Reserve Bank of
Australia.
McCallum, B.T. (1997) Ìssues in the Design of Monetary Policy Rules', NBER Working
Paper, 6016.
McCallum, B.T. and E. Nelson (1999) `Nominal Income Targeting in an Open-economy
Optimizing Model', Journal of Monetary Economics, No. 43, 3.
Neumann, M. (1997) `Monetary Targeting in Germany', in K. Iwao (ed.), Towards More
Effective Monetary Policy, New York, St. Martin's Press.
Onatski, A. and J.H. Stock (1998) `Robust Monetary Policy under Model Uncertainty in a
Small Model of the US Economy', Harvard University, Working Paper.
Orphanides, A. (1998) `Monetary Policy Evaluation with Noisy Information', Federal
Reserve Board, Working Paper.
Orphanides, A. and V. Wieland (1998) `Price Stability and Monetary Policy Effectiveness
when Nominal Interest Rates Are Bounded at Zero', Federal Reserve Board, Working
Paper.
Pearlman, J.G. (1992) `Reputational and Nonreputational Policies under Partial
Information', Journal of Economic Dynamics and Control, 16, 339±57.
Pearlman, J.G., D. Currie and P. Levine (1986) `Rational Expectations Models with Partial
Information', Economic Modelling, 3, 90±105.
Peersman, G. and F. Smets (1998) Ùncertainty and the Taylor Rule in a Simple Model of the
Euro-Area Economy', Working Paper, presented at a conference on Monetary Policy and
Monetary Institutions held on March 5, 1999 under the joint sponsorship of the Federal
Reserve Bank of San Fransciso and the Stanford Institute for Economic Policy Research.
Reifschneider, D.L., D.J. Stockton and D.W. Wilcox (1997) Èconometric Models and the
Monetary Policy Process', Carnegie±Rochester Conference Series on Public Policy, 47, 1±37.
Rudebusch, G.D. (1998) Ìs the Fed Too Timid? Monetary Policy in an Uncertain World',
Federal Reserve Bank of San Francisco, Working Paper.
Rudebusch, G.D., and L.E.O. Svensson (1999a) `Policy Rules for In¯ation Targeting', in
Taylor (1999).
ÐÐÐÐ (1999b) Èurosystem Monetary Targeting: Lessons from U.S. Data', NBER
Working Paper, 7179.
Sack, B. (1988) `Does the Fed Act Gradually? A VAR Analysis', Federal Reserve Board,
Discussion Paper, 98±17.
Sargent, T.J. (1999) `Discussion of ``Policy Rules for Open Economies'' by Laurence Ball',
in Taylor (1999).
Smets, F. (1999) Òutput Gap Uncertainty: Does it Matter for the Taylor Rule?' Bank for
International Settlements, Working Paper.
So
È derlind, P. (1998) `Solution and Estimation of RE Macromodels with Optimal Policy',
Stockholm School of Economics, Working Paper.
So È m, U. (1999a) `Monetary Policy with Uncertain Parameters', Stockholm School
È derstro
of Economics, Working Paper.
ÐÐÐÐ (1999b) `Should Central Banks Be More Aggressive?' Stockholm School of
Economics, Working Paper.
Stock, J.H. (1999) `Making Policies Robust to Model Uncertainty: Comment on ``Policy Rules
for In¯ation Targeting'' by Glenn Rudebusch and Lars E.O. Svensson', in Taylor (1999).
Svensson, Lars E.O. (1996) 'Commentary: How Should Monetary Policy Respond to
Shocks while Maintaining Long-run Price Stability?±Conceptual Issues', in Federal
Reserve Bank of Kansas City (1996).
ÐÐÐÐ (1997a) Ìn¯ation Forecast Targeting: Implementing and Monitoring In¯ation

Targets', European Economic Review, 41, 1111±46.
ÐÐÐÐ (1997b) `Price Level Targeting vs. In¯ation Targeting', Working Paper version.
ÐÐÐÐ (1998a) Ìn¯ation Targeting as a Monetary Policy Rule', NBER Working Paper,
6790.
ÐÐÐÐ (1998b) Ìn¯ation Targeting in an Open Economy: Strict vs. Flexible In¯ation
Targeting', Victoria Economic Commentaries, 15-1, Victoria University of Wellington.
ÐÐÐÐ (1999a) `Better to Respond to Determinants of Targets than to Targets
Themselves', mimeo.
ÐÐÐÐ (1999b) Ìn¯ation Targeting as a Monetary Policy Rule', Journal of Monetary
Economics, 43, 607±54.
ÐÐÐÐ (1999c) Ìn¯ation Targeting: Some Extensions', Scandinavian Journal of
Economics, 101(3), 337±61.
ÐÐÐÐ (1999d) `Monetary Policy Issues for the Eurosystem', Carnegie±Rochester
Conference Series on Public Policy, 51±1, 79±136.
ÐÐÐÐ (1999e) `Price Level Targeting vs. In¯ation Targeting', Journal of Money, Credit
and Banking, 31, 277±95.
ÐÐÐÐ (2000a) `Does the P Model Provide Any Rationale for Monetary Targeting?',
Working Paper, German Economic Review 1, 69±81.
ÐÐÐÐ (2000b) Òpen-Economy In¯ation Targeting', Journal of International Economics,
Vol. no. 50, 155±83.
Svensson, Lars E.O. and M. Woodford (2000) Ìndicators for Optimal Policy under
Asymmetric Information', mimeo.
Sveriges Riksbank (1997), In¯ation Report, September 1997, Stockholm: Sveriges Riksbank.
Taylor, John B. (1996) `How Should Monetary Policy Respond to Shocks while
Maintaining Long-run Price Stability ± Conceptual Issues', in Federal Reserve Bank
of Kansas City (1996).
ÐÐÐÐ (1998) `The Robustness and Ef®ciency of Monetary Policy Rules as Guidelines
for Interest Rate Setting by the European Central Bank', IIES Seminar Paper, 649.
Taylor, John B. (ed.) (1999) Monetary Policy Rules, Chicago, Chicago University Press.
Tinsley, P.A. (1975) Òn Proximate Exploitation of Intermediate Information in
Macroeconomic Forecasting', Federal Reserve Board, Special Studies Paper, 59.
Todter, K.-H., and H.-E. Reimers (1994) `P-Star as a Link between Money and Prices in
Germany', Weltwirtschaftliches Archiv, 130, 273±89.
Vickers, J. (1998) Ìn¯ation Targeting in Practice: the UK Experience', Bank of England
Quarterly Bulletin, 38, 368±75.
von Hagen, Jurgen (1995) Ìn¯ation and Monetary Targeting in Germany', in Leiderman
and Svensson (1995).
Wallis, K.F. (1999) Àsymmetric Density Forecasts of In¯ation and Bank of England's Fan
Chart', National Institute Economic Review, 1/99, 106±12.
Wieland, V. (1996) `Monetary Policy, Parameter Uncertainty and Optimal Learning',
Federal Reserve Board, Working Paper.
ÐÐÐÐ (1998) `Monetary Policy and Uncertainty about the Natural Unemployment
Rate', Federal Reserve Board, Finance and Economics Discussion Paper, 22.
Williams, J.C. (1997) `Simple Rules for Monetary Policy', Federal Reserve Board, Working
Paper.
Wolman, A.L. (1998a) `Staggered Price Setting and the Zero Bound on Nominal Interest
Rates', Federal Reserve Bank of Richmond Economic Quarterly, 84(4), 1±24.
ÐÐÐÐ (1998b) `Real Implications of the Zero Bound on Nominal Interest Rates',
Federal Reserve Bank of Richmond, Working Paper.
Woodford, M. (1994) `Nonstandard Indicators for Monetary Policy: Can Their

Usefulness Be Judged from Forecasting Regressions?' in G.N. Mankiw, (ed.), Monetary
Policy, Chicago, University of Chicago Press.
Discussion
Mervyn King
It has been said about a great newspaper that àll human life is there'. It could
equally be said about this chapter that àll Lars Svensson is there'. What
Chapter 2 does is to bring together many of Svensson's contributions to the
analysis of in¯ation targets, together with some new ideas, and ask the
question ± what can the European Central Bank (ECB) learn from this work? It
is a good question, and I hope that even if the ECB does not agree with
everything that Lars says, it will nevertheless bene®t from a serious discussion
of the points which he makes. I should like at the outset to acknowledge the
personal contribution which Lars Svensson has made to the theoretical
analysis of in¯ation targets. Those of us working in central banks with
in¯ation targets know only too well the debt we owe him for the intellectual
development which he has made. And, most important, this has been made in
a spirit of intellectual open-mindedness. This is important. No central bank
can claim a monopoly of wisdom, and in¯ation targets share much in
common with more traditional monetary policy frameworks such as monetary
targets. Both approaches can learn from the other. But it is clear that the 1990s
have been the decade of in¯ation targets, and Svensson has made a major
contribution to that achievement.
The central point in Svensson's chapter is the need for à systematic
operational framework for policy decisions by central banks'. Svensson sees
forecast targeting as the most sensible way forward. Among the class of policy
frameworks in this category in¯ation targeting appears superior. Svensson
contrasts the systematic approach of forecast targeting with two alternatives.
These are, ®rst, simple instrument rules and, second, intermediate targets. I
agree with Svensson's ranking of these three alternative approaches to
monetary policy. But one must be a little cautious before drawing hard and
fast distinctions between them. No advocate of instrument rules would
suggest that policy be put on autopilot and that the rule be used in a
completely mechanical way to set policy. Equally, advocates of intermediate
monetary targets would wish, in certain circumstances, to deviate from the
policy implied by a mechanical application of those targets, as the Bundesbank
103
104 Mervyn King
did with money targets for many years. Of course, if the deviation from the
rule or the target is too great then its use either as a means of discipline or a
form of communication becomes low. But there is equally no simple
mechanical link between any particular summary statistic of the in¯ation
forecast and the choice of the policy instrument (usually the short-term
interest rate). There is always a judgement about what policy setting is
appropriate given the outlook for in¯ation.
The reason I prefer in¯ation targeting is that it makes clear that the
discretion used is in the mapping from the outlook for in¯ation to the choice
of instrument. That enables the central bank to divide its work into two parts.
The ®rst is to think through and then explain its assessments of the outlook for
in¯ation and output implied by the current state of the economy. The second
is to explain why, given that outlook, a particular policy choice was
appropriate. That seems to me to correspond to what central banks do in
practice. In contrast, simple instrument rules can be useful cross-checks on
choices made but are not very useful as communication strategies. And
intermediate targets likewise pose the problem of how one explains a
deviation of policy from that implied by the intermediate target. Surely that
would end up being an explanation in terms of the outlook for in¯ation.
Hence I like Svensson's phrase `distribution forecast targeting'.
But it is important to note that this is a framework for policy and not an
optimal rule for interest rates. And one factor which is striking in practice is
that the choice between the three approaches identi®ed by Svensson implies
nothing about views concerning the transmission mechanism of monetary
policy. Indeed, I could quite easily envisage my òptimal monetary policy
council', whether for the ECB or any other area, comprising individuals who
differed in their belief about the choice of policy framework. What matters
most is their ability to think of the transmission mechanism and to form
judgements about the current state of the economy.
Let me turn now to four speci®c points on Svensson's chapter. First,
Svensson advocates moving gradually from in¯ation targets to price level
targets. I have some sympathy with his arguments. There is still the important
question of how long is the horizon over which the price level is brought back
to its target path. It would be useful to carry out simulations to judge how far
an instrument rule, such as the Taylor rule, would imply a different interest
rate path if an error-correction term were added in order to bring the price level
back to a predetermined path within some reasonable time horizon. How
much greater would the volatility of interest rates be? But there is also a deeper
question of whether the costs of in¯ation are more to do with instability of the
expected price level over long horizons or of in¯ation itself. More work is
required on that question which, understandably, is not the focus of
Svensson's chapter.
Second, Svensson criticises the ECB for announcing a range for their
de®nition of price stability (close to an in¯ation target) rather than a point
Discussion 105
target. Again, I have some sympathy with this point. The symmetry of the
Bank of England's in¯ation target has been an immensely helpful aspect of our
remit because we have been able to show that it can be necessary to increase or
decrease interest rates according to circumstances. It is important to avoid
giving the impression that central banks are reluctant to cut interest rates and
quick to raise them. That would be the opposite of the politician's preference
for raising interest rates `when necessary' and lowering them `when possible'.
Symmetry should be an important part of the approach to setting interest
rates. The problem with a target range is that there is a lack of clarity about the
objective of monetary policy. Of course one might argue that the central bank
would do best by aiming for the centre of the range, and economic agents
would know that. The analogy is that, faced with an open goal and with some
uncertainty about your ability to kick a football, the best advice is to aim for
the point between the middle of the goalposts. Sadly, many footballers of my
acquaintance seem to be confused by an open goal, hesitate, and too often
miss. A point target has the virtue of concentrating the mind. It is also
particularly useful in the context of a policy making committee, to
demonstrate that the concept of `hawks' and `doves' on the same committee
has no meaning when the committee has a common point in¯ation target.
Third, Svensson suggests that the use of the reference value for money by
the ECB is misguided. The announcement of a money target might confuse
agents rather than ànchor' in¯ationary expectations. I have more sympathy
with the idea that money is special. There is still much that we do not fully
understand about the transmission mechanism, but it is dif®cult to talk about
in¯ation or monetary policy without according a special role for money. And
many of the models which play down the role of money and give a prominent
place to such concepts as the neutral real interest rate or the equilibrium
exchange rate presume more knowledge about the empirical values of these
concepts than is available in practice to central banks. So I think there are two
reasons for allowing a special role to money. The ®rst is that given uncertainty
about the appropriate model for the transmission mechanism, money may
have a robust property with respect to nominal outcomes. Second, there is a
real danger that focusing solely on traditional econometric models results in
trying to explain in¯ation solely in terms of real variables. So it is extremely
important that a central bank sees as one of its ®rst responsibilities the need to
explain movements in money and why it thinks that its policy stance is
consistent with the likely behaviour of money.
Fourth, Svensson considers the choice between targeting the mean, median
or modal future in¯ation rate. He points out that different loss functions
might imply the need to target different summary statistics of the distribution
of future in¯ation. More generally, one of the aspects of central bank
discretion is to decide how it should react to the entire distribution of in¯ation
prospects rather than commit itself to targeting a particular summary statistic
in all circumstances. Hence I much prefer Svensson's approach of considering
106 Mervyn King
the distribution of in¯ation as a whole and not just summary statistics such as
the mean, median or mode. That is also why in the `fan charts' for the in¯ation
outlook which are published in the Bank of England's In¯ation Report, we stress
that there can be no mechanical link between any particular line on that chart
and the policy decision. In some circumstances it may make sense to focus
very much on the mean outlook, and there is no doubt that over a long period
any deviation from that must cancel out if the target is to be met on average.
But there may be circumstances in which the mean is affected by a very small
probability of a signi®cant upside or downside risk. And it may not make sense
to adjust policy today to that eventuality. The fan chart should be constructed
in order to present information about the future outlook for in¯ation ± the
entire distribution ± in an informative way. The policy decision is separate.
In conclusion, I would stress that all central banks, whatever policy
framework they use, face essentially the same problem. The lags in monetary
policy mean that they must look forward to assess the impact of policy
decisions on the outlook for in¯ation and output. There are important
differences in the way in which the policy decision is presented, and the
choice of framework is not unimportant. As Svensson, I prefer the in¯ation
target approach. But I do believe that central banks can learn from each other,
and that just as the Bank of England learnt a great deal from the experience of
the Bundesbank in the period when it was developing its new approach and
acquiring the credibility which was necessary for the granting of independ-
ence in 1997, I hope that our experience will offer some lessons for the ECB.
Discussion
Jose Vin
Äals
Chapter 2 is a work in the best Svensson tradition, in the sense that it

summarises, completes and builds on previous work by the author in the ®eld
of monetary policy; it is written in a very clear, organised and pedagogical
manner; and it makes use of economic theory to obtain results which are of
direct policy relevance in general and also for the Eurosystem.
The chapter addresses three blocks of issues related, respectively, to the
de®nition of price stability, the maintenance of price stability and the lessons
to be drawn for the Eurosystem. Although I shall say something on each of
these issues, my comments will focus mainly on the second issue, which I
consider to be the most important part of the chapter.
On the de®nition of price stability, I believe that the main question addressed
by the author is whether price stability should be understood as constancy of
the price level or as a suf®ciently low rate of in¯ation. From a conceptual
viewpoint, the debate is far from settled, and it is presented as such in the
chapter. Still, I think it is interesting to consider Svensson's idea that constancy
of the price level may not only be preferable on long-term grounds ± since it
leads to a lower variance of the price level ± but, under certain circumstances,
also on short-term grounds. In particular, in those cases where there is a serious
risk of de¯ation or when de¯ation actually occurs, having a constant price-level
objective would be preferable to having a low in¯ation objective as long as it is
credible before the eyes of the public. This would make in¯ationary
expectations higher, as a result of the expected return of the price level to its
initial value, and real interest rates considerably lower. More generally, however,
as the author himself recognises, the issue of whether a constant price level or a
suf®ciently low rate of in¯ation is a preferable de®nition of price stability is far
from being conceptually settled. For this reason, he resorts to a pragmatic
solution to my liking: for the time being, achieving and maintaining low
in¯ation is a suf®ciently ambitious objective for price stability. This is,
moreover, the solution chosen by the Eurosystem and other central banks.
The discussion of the issues concerning the maintenance of price stability is,
in my view, the core of the chapter and where the most substantive points are
107
108 Jose Vin
Äals
raised, both from a theoretical and a practical perspective. Svensson compares

the relative merits of three strategies for monetary policy: commitment to a
simple instrument rule (e.g. a Taylor rule or a money base rule), forecast
targeting (e.g. in¯ation-forecast targeting) and intermediate targeting (e.g.
monetary targeting).
Before going any further, I would like to say that if the author had written
his paper ten years ago, he would have had an even more dif®cult task than at
present in trying to ®nd out which is the best way of conducting monetary
policy ± that is, which is the best monetary policy practice. The reason why
this task is made somewhat easier nowadays ± without having at all become
trivial ± is that in recent years a broad consensus has emerged both among
economists and central bankers about what should be the most appropriate
institutional framework for monetary policy. Consequently, whatever the
strategy chosen, monetary policy decisions are nowadays taken within an
institutional framework where the maintenance of price stability is the main,
the primary or, at least, an important goal of monetary policy, and where
central banks enjoy a considerable degree of independence in the pursuit of
such goal.
Without going as far as saying that this makes which strategic framework is
®nally chosen to conduct monetary policy irrelevant, it nevertheless cannot
be denied that those central banks which pursue price stability with a
considerable degree of independence, in spite of adhering to formally very
different monetary policy strategies tend, in practice, to look at broadly
comparable sets of economic variables ± albeit in more or less formal ways ±
and to make interest rate decisions that are rather similar when faced with
similar circumstances. Yet, it is of paramount importance to realise that
alternative strategies are not equally effective in reaching the established
policy objectives. Thus, it becomes indispensable to carry out a comparative
analysis of the available strategies, such as that performed by Svensson in the
chapter 2.
Before critically appraising the analysis, let me just point out what I would
personally view as desirable features of what constitutes 'best monetary policy
practice'. First, it should combine rigour and discipline in the medium-term
policy stance with some room for ¯exibility to respond to certain macro-
economic disturbances in the shorter term. Secondly, it should be forward-
looking, given the lags with which monetary policy affects price develop-
ments. Thirdly, it should not be run mechanistically, but rather make an
ef®cient use of the available information, both model-based and extra-model.
Fourthly, it should ensure that the decision-making process performs reason-
ably well across different models of the economy, given how imperfect our
knowledge is about the transmission mechanism of monetary policy. And
®nally, both the strategy and the decisions made on the basis of it should be
clearly and transparently communicated to the public, so as to favourably
in¯uence the private sector's expectations.
Discussion 109
In principle, such 'best monetary policy practice' ought to be able to solve

the problems frequently encountered in the rules vs. discretion debate. On the
one hand, it would avoid the dynamic consistency problems arising from an
excessively discretionary management of monetary policy, which make it
impossible to maintain price stability. And, on the other, it would overcome
the drawbacks associated with ®xed or rigid rules which, by predetermining
the course of monetary variables or instruments, may unduly constrain, in
actual practice, policy-makers' scope to react to useful information.
The comparison of alternative strategic frameworks for maintaining price
stability made by the author is based on a model where the central bank
optimises its loss function subject to the restrictions implied by the structure
of the economy. His analysis yields the result that the preferable monetary
policy strategy is what he terms 'forecast targeting', since it is considered to be
superior both to what he calls 'simple instrument rules' and 'intermediate
targeting'. While I share the thrust of the author's basic conclusions, I would
like to discuss his analysis from the standpoint of the considerations I made
earlier on. Indeed, I believe that part of the superiority of 'forecast targeting' is
much more due to easiness of communication within the central bank and
also with the public than Svensson actually acknowledges.
Concerning the comparison between instrument rules and forecast
targeting, a crucial distinction is drawn between the òptimal' instrument
rule and a `simple' instrument rule. I ®nd the arguments in the chapter against
the commitment to a `simple' instrument rule fairly convincing. While it has
been argued by some authors that simple rules may not be optimal in any
given model but nevertheless can be superior in the sense of delivering
reasonably good results across models, I agree with Svensson that, in general,
simple instrument rules are too rigid to be valid in a real world monetary policy
decision framework. The reason is that they prevent judgmental analysis and
extra-model information from being properly taken into account, not to
mention problems of incentive compatibility.
Nevertheless, as pointed out in the chapter, there is also an òptimal'
instrument rule, which is the solution to the general optimisation problem
faced by the central bank. Theoretically, this is neither better nor worse than
forecast targeting since both are alternative ways of characterising the optimal
monetary policy. Consequently, provided that any preferred alternative to an
optimal instrument rule must also ful®l the requirement of being a solution to
the general optimisation problem, the reasons why the optimal instrument
rule is not considered by the author to be appropriate merit careful attention.
Interestingly, he disregards the optimal instrument rule because, on the one
hand, it is very sensitive to the underlying model of the economy ± which in
practice is not very precisely known ±, and, on the other, because it is overly
complex as a policy guide. As concerns the ®rst reason, I wonder whether
model uncertainty does not have the same unfavourable consequences for
forecast targeting as for instrument rules, and whether such complications
110 Jose Vin
Äals
could be taken into account in an analogous manner as with forecast

targeting, the latter being explained in detail in the chapter. Insofar as this
could be done ± and I think it could ±, then the only remaining argument
against the optimal instrument rule would be its complexity and thus the
dif®culty in fostering ready communication with the public. If, because of
such reasons, the optimal instrument rule is discarded in favour of a simpler
rule, then forecast targeting would be superior. The obvious reason is that
the simpler rule will not be exploiting appropriately the available
information, in addition to the other problems of incentive-compatibility
mentioned earlier on.
Regarding the comparison between forecast targeting and intermediate
targeting, insofar as we restrict ourselves to the conceptual setting used by the
author, it follows that only under very restrictive and well known conditions ±
i.e. a recursive macroeconomic structure ± will intermediate targeting be as
good as forecast targeting. In fact, even those central banks adhering formally
to intermediate targeting ± like monetary targeting ± have pursued their
strategies in a rather pragmatic manner, thus perhaps proving how unlikely it
is that the conditions for pure intermediate targeting to be the 'best policy
practice' are satis®ed in the real world.
Overall, I think that Svensson's defence of forecast targeting as
constituting the `best monetary policy practice' can be also rationalised
in terms of the ®ve desirable features that such practice should contain and
which I alluded to earlier in my discussion. In this regard, forecast targeting
would yield a good mix between medium-term rigour and short-term
¯exibility; it would constitute an organised framework for systematic
decision making inside the central bank which makes the best possible use
of all the available information; it would be forward-looking and non-
mechanistic in the sense of comparing forecasts across various models of
the economy and using extra-model information to arrive at the best
possible overall judgement on prospective price developments; and ®nally,
it would make it easier for the public to understand monetary policy
decisions by placing the focus of attention on the key variable, namely
price developments.
Indeed, any central banker knows that, in practice, monetary policy
decisions are taken in a forward-looking manner, something that becomes
necessary given the lags with which monetary policy impacts the economy
and, in particular, prices. Therefore, decisions are made, explicitly or
implicitly, on the basis of the best appraisal of the price outlook, something
which is nothing else but Lars Svensson's `best conditional forecast'.
But if so, where do the differences lie among monetary policy strategies? In
my view, in the following three domains. First, in the extent to which the
forecast is or not an explicit feature of the strategy, which serves to determine
whether policy is formally conducted according to a forecast targeting scheme
or not. Secondly, in the extent to which the ± explicit or implicit ± forecast is
Discussion 111
based on macroeconometric models or other less formal tools. And ®nally, in

the extent to which the conditional forecast is published and thus serves as an
explicit guide for policy action in the eyes of the public.
To conclude, as Alan Greenspan has stated 'There is no way to avoid making
a forecast, explicit or implicit, when making monetary policy'. The issue is
whether, as Lars Svensson argues, forecast targeting is the best way for central
banks to make use of the available information in the most ef®cient possible
way to maintain price stability. The experience of those central banks which
have adopted such a framework in the past, although such experience is still
relatively short compared to other longer-standing strategies, suggests that
forecast targeting is worth considering.
3
The Transmission Process1

Allan H. Meltzer
I must confess some vested interests in this topic. I ®rst discussed it in a paper
with Karl Brunner, more than thirty-®ve years ago (Brunner and Meltzer,
1963). I have returned to the topic many times, most recently in a published
symposium (Meltzer, 1995). I will refrain from reviewing these earlier studies,
although I will refer in passing to some of the main ideas. I will concentrate on
two topics. They do not exhaust the subject, but they raise issues that I believe
are central.
First, I raise some issues about a current class of models of monetary
transmission in which a short-term interest rate represents the transmission
process. The class of models is so widely accepted that the conclusions I
challenge have become part of the canon. A different class of models ± a more
useful one, I believe ± does more than give different answers. Some issues do
not arise; they are no longer relevant. And some issues remain relevant but
receive a different answer. The role of money is one such issue.
Second, I discuss some of the evidence I have gathered from my study ± A
History of the Federal Reserve ± the work that has been my main occupation for
the past four years. The two pieces are related, as I hope to show. The evidence
from history shows that the transmission process cannot be summarised by a
single interest rate. In the ®nal section, I present some econometric evidence
to supplement the historical data.
1 Standard models of monetary transmission
The core or central issue about the transmission of monetary policy is: How
does a monetary impulse affect relative prices and real demands as it moves
through the economy from its ®rst appearance to its ultimate effect on the
main determinants of economic welfare? The standard answer to this question
is, now as for some considerable time in the past, that monetary injections
change an interest rate in the money market. Because some prices, money
wages, or anticipations do not adjust instantly, there are effects on output and
employment that, though temporary, may be large and costly. Eventually,
112
The Transmission Process 113
relative prices are restored, so the price level changes equi-proportionally.

Money is neutral in the long run.2
Most of us accept large parts of this story. Real business cycle theorists would
both add and omit, as would other economists. Loan or credit market models
make other changes. I shall not dwell on these quali®cations, or the relevance
of these analyses. Instead, I will use as illustrative a model that represents some
of the best recent work on monetary policy to give speci®city where it is
needed.3 Most of my comments apply more generally.
The class of models has two equations, aggregate demand and supply.
Aggregate supply is a Phillips-type curve. Important for current purposes is
that monetary policy is represented only by the interest rate in the aggregate
demand function. An expansive monetary action lowers the nominal interest
rate; a contractive action raises it. The action is unanticipated, at least as to
timing, so the real rate of interest changes with the nominal rate absolutely
and relative to the anticipated real rate. In Svensson's (1998) model, that I
have used as a good, clear example, aggregate demand is replaced by the
output gap, and the aggregate supply function is adjusted accordingly. This
difference is not consequential for my discussion.
The story about transmission embedded in this model is that a change in
interest rates changes aggregate demand in the opposite direction. If the interest
rate increases, output falls, the output gap increases, so in¯ation declines,
raising the real rate of interest and continuing the process of mutual adjustment
of aggregate demand and supply. When the target rate of in¯ation is reached,
the central bank rests, and the economy adjusts to the new equilibrium.
The money stock and the demand for money remain in the background.
The central bank sets the interest rate and provides the stock of money
required to satisfy demand. The price level is determined with the rate of
in¯ation and output, and so too is the demand for real money balances.
This stripped-down model is useful for illustrating the interaction of
aggregate demand and supply. As a model of monetary transmission, I believe
it is de®cient for two main reasons. First, it has implications that are either
misleading or wrong. Second, it omits signi®cant parts of the adjustment of
relative prices and real wealth to monetary and other shocks and the responses
of aggregate demand and output to changes in relative prices and real wealth.
These omissions are major parts of the transmission process.
Two false implications

Many economists have used a model of this kind to discuss monetary policy in
the 1930s depression and the long Japanese recession of the 1990s. In both
cases, short-term nominal interest rates approached zero. The implication
drawn from these experiences is that the monetary authority was powerless
because the short-term interest rate was close to zero. The economy was said to
be in a liquidity trap. (For the Japanese case, see Krugman, 1998, and Ito,
1998).
114 Allan H. Meltzer
A `liquidity' trap is de®ned as a condition in which the demand for nominal

money balances equals the stock of nominal balances at unchanged prices and
interest rates, for all values of the nominal stock. Additions to the nominal
stock cannot be transmitted to the real sector or the price level because the
interest rate has reached a ¯oor at or near zero.
As is customary in this class of models, assume that all assets are gross
substitutes in portfolios. Ignoring transaction and information costs, it does
not matter which asset the central bank buys or sells. Whichever asset it buys,
the market adjusts the relative prices of all assets to re¯ect the change in
relative supplies. This is a powerful and useful argument with strong
implications. For example, economists use it to show that sterilised foreign
exchange market intervention has no effect or to show why unsterilised
exchange rate intervention differs from sterilised intervention.
Suppose, now, that with a short-term interest at zero, the Bank of Japan
announces that it wants the dollar exchange rate to fall by 50 per cent and that
it is prepared to print yen to buy dollars until that occurs. Is there any doubt
that the yen would depreciate or that the depreciation would affect spending,
output and prices in Japan?
Suppose, instead, that the Bank of Japan makes no announcement but buys
dollars with the intention of depreciating the yen by 50 per cent. There may be
differences in the timing of responses, but the ultimate effect would be the
same: monetary expansion would affect the economy. There would be no
liquidity trap whatever the short-term interest rate in the market in which the
central bank usually operates.
Two questions occur. First, how can we reconcile our standard assumption
of gross substitution with this obvious contradiction. Second, does this
argument imply that a liquidity trap is impossible in a multi-asset world?
The liquidity trap, by assumption, makes short-term Treasury bills (or
similar securities) a perfect substitute for base money or bank reserves.
Exchanging one for the other does nothing of interest. Exchanging either
money or Treasury bills for some other asset such as foreign money, domestic
or foreign long-term bonds, equities, or commodities changes relative prices
and real wealth. In this hypothetical case, base money plus bills is a composite
good. The composite good is a gross substitute for other assets; increasing
either component, or both, is expansive.
For a full liquidity trap to be effective, the composite asset ± money plus bills
± must be a perfect substitute for all other assets. When the marginal rate of
substitution of money for bonds goes to zero, all marginal rates of substitution
must go to zero.4 All assets are parts of a single composite good.
If assets other than bills and money remain gross substitutes, a liquidity trap
means only that one row and one column in the matrix of marginal rates of
substitution has been eliminated. All other marginal rates of substitution
remain. Monetary policy remains effective. The standard class of models gives
the wrong answer about policy. It implies that a liquidity trap is possible and,
for some, is a reality, (Krugman, 1988, Ito, 1998). The alternative denies that a
liquidity trap is possible except in the limit when all prices are zero.
A closely related proposition has received considerable attention as in¯ation
rates fell and remained low in the 1990s. Summers (1991) revived the
argument that a zero in¯ation target is socially costly because it sets a lower
bound for nominal interest rates. Monetary policy becomes weak or powerless;
it cannot lower the short-term nominal rate or prevent falling prices from
raising the real rate of interest. With money wages in¯exible downward,
unemployment rises. Akerlof, Dickens and Perry (1996) perform the remark-
able feat of ®nding evidence for this proposition using data for a period in
which in¯ation never remained close to zero. And Benhabib, Schmitt-Grohe
and Uribe (1998) argue that it is perilous to use a Taylor rule when in¯ation is
near zero.
A more sophisticated version of Summers' argument uses a stochastic model
with non-linearity in the transmission process when in¯ation is below 2 per
cent. Orphanides and Wieland (1998) ®nd that there is no evidence of an
operative lower bound in US postwar data. They claim that the lower bound
was in effect during the 1930s, so monetary policy was in¯exible for part of
that decade.
For this claim to be true, the short-term interest rate must be the principal or
only means by which monetary actions are transmitted from the central bank,
through the market, to the economy. As my old friend Karl Brunner often said:
we know this is false. Monetary actions are effective and powerful in the less
developed countries of Africa, Latin America, or Asia where there is no money
market. Relative prices respond to monetary impulses in countries without
central banks, and without money markets. There is more to the transmission
process than the models recognise.
Do these additional channels operate in countries with active money
markets? In writing the History of the Federal Reserve, I have found three
relevant examples from the years 1914 to 1951 that I have researched to date.
2 Historical evidence
Each of the examples I consider concerns a period of de¯ation. Prices fell,

raising real interest rates through most or all of the recession. Expansive ®scal
actions in each episode were either modest or absent. Two of the recessions are
considered severe according to rankings by the National Bureau of Economic
Research. In each case the economy recovered, and two of the three recessions
were of average length.
The common feature that is relevant for the monetary transmission process
is that real money balances and real interest rates rose together. In each case
there was a common cause: prices fell. In some cases gold in¯ows or Federal
Reserve actions increased the monetary base; in other cases, the monetary
authorities were passive or restrictive through most or all of the recession.
Differences of this kind are of secondary importance in the three examples

(but not in the Great Depression). The dominant, common impulse in the
three examples was de¯ation.
Two of the three episodes share a second relevant feature: the interest rate
on short-term Treasury bills was historically low. During the 1948±9 recession,
rates on Treasury bills were about one per cent. In 1937±8, bill rates were close
to zero. In the third case, 1920±1, short-term rates remained well above zero,
but the de¯ation was sharp and severe, so real interest rates and real money
balances rose together.
1937±8
The National Bureau ranks the 1937±8 recession as the third most severe
recession in the years after the First World War. Real GNP fell 18 per cent and
industrial production 32 per cent in the thirteen months from May 1937 to
June 1938. Unemployment reached a peak of 20 per cent, not very different
from the 25 per cent peak in 1932.
The probable causes of the recession include both ®scal and monetary
actions. There is a very large reduction in the government de®cit in 1937 and a
very large reduction in growth of the monetary base. The main ®scal actions
are the end of the soldiers' bonus payment, the enactment of an excess pro®ts
tax to pay for part of the bonuses in ®scal 1937, and the start of social security
tax collections in ®scal 1936. The soldiers' bonus is the largest item,
$1.7 billion of current spending. It was paid in June 1936, in time for the
election later that year. The bonus was paid in bonds, but the bonds could be
sold for cash. By December 1936, $1.4 billion had been cashed. Balke and
Gordon's (1996) quarterly data show an 18 per cent average rate of increase in
real GNP for the last three quarters of 1936.
The most important monetary actions are the beginning of gold sterilisation
at the end of 1936 and the second and third increase in reserve requirement
ratios in March and May 1937. These increases completed the doubling of
reserve requirement ratios between August 1936 and May 1937. During the
entire period December 1936 to December 1938 that brackets the recession,
interest rates on Treasury bills remained between 0.03 per cent and 0.56 per
cent. Long-term nominal rates on Treasury bonds were modestly higher
during the recession than before or after, but the difference is small; the range
is 2.55 per cent ± 2.83 per cent.
Annualised monthly rates of price change are consistently negative from
October 1937 to February 1938 and intermittently negative for the rest of
1938. To smooth the data, I used moving twelve month averages of rates of
price change. Figure 3.1 compares the real interest rate to the annual growth
of the monetary base.
The common element in the two series is the twelve-month moving average
of the rate of price change. The divergence between the two series re¯ects some
release of sterilised gold into the monetary base in September 1937 and a small
per cent per cent
+7.5 +25
+6.0 +20
+4.5 +15
Year-over-year real base growth

Real long-term interest rate
+3.0 Real long-term interest rate +10
+1.5 +5
0 0
–1.5 –5
–3.0 Year-over-year real base growth –10
–4.5 –15
6.0
–6.0 –20
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
1936 1937 1938
117
Figure 3.1 Year-over-year real base growth versus real long-term interest rate, January 1936±December 1938
volume ($38 million) of open market purchases in November, principally for

seasonal reasons.
Not until February 1938, after nine months of a deep recession, did the
Federal Reserve propose countercyclical action, the release of additional gold
from sterilisation. In April, the Roosevelt Administration announced
$2 billion of additional government spending for construction and relief. As
part of this programme, the Treasury released another $1.4 billion from
sterilisation and the Federal Reserve released $750 million of reserves by
lowering reserve requirement ratios. Figure 3.1 shows the sustained rapid
increase in the real value of the monetary base beginning in February 1938.
Real ®nal sales rose in the following quarter, but inventories fell, so real GNP
did not increase until the third quarter.
What does this episode suggest about the transmission of monetary
policy? In the months preceding recovery, and in the early months of
expansion, the real interest rate rose from 2.9 per cent in January 1938 to
more than 6 per cent in September±November 1938. Although nominal rates
remained historically low, real rates were relatively high. In contrast, real
money balances accelerate ®ve months before the end of the recession:
between February and June, growth of real balances rises from ±7.6 per cent
to 17.6 per cent. By the end of 1938, growth of real balances reached an
almost 25 per cent annual rate.
I draw three conclusions for the transmission process. First, low nominal
interest rates misled the Federal Reserve, on this occasion as on others, into
believing that monetary policy was expansive. Second, although short-term
interest rates stayed at or near zero, monetary policy was not powerless.
Desterilising gold to increase the monetary base raised nominal and real
money balances and increased spending. Third, the ®nancial system was not
in a liquidity trap. Channels other than the short-term interest rate
transmitted monetary expansion to output and the price level.
1948±9
The 1948±9 recession provides a second example refuting the liquidity trap
and the small or vanishing effect of monetary policy at low nominal interest
rates. The Federal Reserve pegged nominal long-term interest rates below the
2.5 per cent ceiling in effect from 1942 to 1951. Despite the pegging policy,
the monetary base fell through most of 1948. The principal reason is that the
Treasury used its budget surplus to retire debt held by the Reserve banks.
The monetary base fell as a consequence of the Treasury's actions. Although
the Federal Reserve complained about being an engine of in¯ation, prices fell
in half the months of 1948 and 1949.
The National Bureau dates the end of the expansion in November 1948 and
the recession trough in October 1949. The twelve-month moving average rate
of in¯ation fell from above 9 per cent in June and July 1948 to negative values
in May 1949. It remained negative for the rest of that year.
During most of the recession the Federal Reserve was more concerned about
a return of in¯ation than about the recession. The nominal rate on Treasury
bills remained between 1.02 per cent and 1.17 per cent throughout the
recession. Figure 3.2 compares annual growth of the real monetary base to the
real interest rate in the two years that include the recession. Data are computed
as in Figure 3.1. As before, the high positive correlation re¯ects the common
effect of the rate of price change on the two series. The high correlation and
parallel movement show that until late in 1949, when the recession was
almost over, the Federal Reserve took few actions to increase base growth.
Real base growth fell to ±11 per cent in September 1948, two months before
the cyclical peak. Thereafter base growth rose, but did not become positive
until April 1949, six months before the trough. The peak rate of base growth is
close to 6 per cent in August 1949, two months before the end of the recession.
At that time the real long-term interest rate was above 5 per cent.
Once again, the movement of real base growth is consistent with the
beginning and end of recession; the movement of real interest rates is not.
Once again, low nominal short-term interest rates do not appear to have
weakened the effects of monetary policy. And once again there appears to be
more to the transmission process than is contained in standard models.
1920±1
The third episode is the recession from January 1920 to July 1921.5 The
National Bureau ends the expansion in January 1920 and puts the last month
of recession in June 1921. The Federal Reserve undertook larger policy actions,
so nominal interest rates and nominal base growth re¯ect these actions.
In¯ationary policies in much of Europe and restrictive policies in the United
States brought an in¯ow of gold. The base and interest rate changes re¯ect
these in¯uences also.
Nevertheless, real base growth and real interest rates are positively correlated
during the recession. Both are negative at the start of the recession, turn
positive about a year later and reach a peak at the end of the recession. Judged
by base growth, monetary actions are countercyclical in the ®rst half of 1921;
judged by real interest rates, these actions are procyclical.
Figure 3.3 shows these data. The long-term nominal rate remains within a
narrow range but is higher at the trough of the recession than at the previous
peak. The dominant in¯uence on real rates and real base growth during the
recession is the decline in in¯ation followed by de¯ation.
As in the previous two episodes, interest rates give a misleading signal about
the thrust of policy. Real base growth gives a more correct signal. In this
recession, the de¯ation is severe; the peak annualised rate reached 17 per cent,
and it was above 10 per cent for ten consecutive months. The real long-term
interest rate, (i ± )/(1 ), is above 25 per cent at the end of the recession. The
economy recovered despite, not because of, the level of real interest rates.
120
per cent
per cent
+4.5 +6
month moving avarage real base growth

+3.0
Real long-term interest rate +4
+1.5 +2
0 0
–1.5 –2
–3.0 –4
–4.5
Twelve-month
–6
Twelve-month moving avarage real base growth
–6.0 Twelve month moving avarage real base growth –8
Twelve
–7.5 –10
–9.0 –12
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
1948 1949
Figure 3.2 Twelve-month moving average real base growth versus real long-term interest rate, December 1947±December 1949
per cent
per cent

Real long-term interest rate +30 +10
Year- over- year real base growth

+15
+5
0
0
–15
Year-over-year real base growth –5
–30 –10
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
1920 1921 1922
Figure 3.3 Year-over-year real base growth versus real long-term interest rate, January 1920±December 1922
121
The three historical periods raise doubts about the central role assigned to
interest rates in the transmission process. They suggest an important role for
real balances. I return to these issues below.
The Great Depression

The only other period of US de¯ation after 1914 is during 1929±33, the Great
Depression. The real interest rate rose from 5 per cent to 15 per cent and
remained near 15 per cent through the last two years of the recession. Real base
growth rose once bank runs began late in 1930, but this is, of course,
misleading. As Figure 3.4 shows, growth of real balances ± measured here by
M1 ± is very different in this period than in other de¯ationary periods. A
principal difference is that monetary contraction was strong enough to offset
the effects of de¯ation on real balances.
An additional problem
Although the central bank's objective function is not directly part of the
transmission process, the three US recessions considered here raise doubts
about the choice of objective function used in current analyses. The usual
function has two arguments: (1) the difference between the in¯ation target
and the actual or expected in¯ation rate and (2) the output gap, the difference
between actual output and output consistent with the natural rate. Yet, we
have seen that faced with de¯ation and a deep recession (in three of the four
episodes) the Federal Reserve was slow to act. And it did not act effectively to
end de¯ation and recession. The objective function fails in these cases as a
positive statement of central bank objectives.6
My concerns about the standard objective function are not limited to its
empirical support, important as that is. I believe that using the output gap as
an objective of the central bank is problematic. This gap can arise for reasons
unrelated to monetary policy actions ± for example, an oil shock, reductions in
employment and output in the European Union resulting from provisions of
the welfare state, or other real events. My colleague Bennett McCallum
suggests that the problem can be overcome by rede®ning the natural rate to
take account of non-monetary effects (Chapter 1 in this volume). In principle,
this can be done; in practice it is dif®cult to do accurately.
The natural rate is not like the gravity constant. We neither measure it
precisely nor agree on its value. Opening the objective function to the pull and
tug of opinions about the size of the gap carries a risk to economic stability and
the apolitical position of an independent central bank. Recent discussion in
Europe and the United States shows that there can be differences of opinion
about measurement of both potential output and the natural rate of
unemployment.
A second problem with the now standard objective function is that it
neglects several issues of concern to central bankers. There is no role for a
lender of last resort. A run to currency, bank failures, a drain of foreign
per cent per cent
Real interest rate

+15 +15
+10 +10
+5 +5
0
0
–5 Real M1 growth –5
–10 –10
Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1
1929 1930 1931 1932 1933
Figure 3.4 Real M1 growth and real interest rate, August 1929±March 1993
123
exchange reserves, a major change in the exchange rate, or a so-called `credit

crunch' enters the bank's objective function only by changing the output gap or
in¯ation. This misstates what central banks do. Properly functioning central
banks respond to the increased demand for money before output and prices fall.
Of®cials of the new European Central Bank (ECB) emphasise that in¯ation is
their sole objective. Their insistence on this point does not, and should not,
preclude concern for the cost of achieving the zero or target rate of in¯ation.
This cost should appear in the objective function, so that the objective is not
just to achieve zero in¯ation but to do so at minimum social cost.
With cost replacing the output gap in the objective function, the way is
open to treat other issues of concern. The implicit cost function used by
central banks includes costs of maintaining or restoring ®nancial safety or
solvency, avoiding a credit crunch, or increasing unemployment. In practice,
even central banks most concerned about in¯ation do not ignore the social
cost of reducing in¯ation. They proceed gradually, over a period of months or
years. And they, properly, do not ignore ®nancial fragility, ®nancial failures
and lender of last resort responsibilities when making decisions. In fact, the
opposite seems to be true: some central banks have recently been overly
sensitive to ®nancial fragility. The Federal Reserve's decisions to reduce
interest rates three times in the autumn of 1998 suggests that ®nancial
fragility, or its prospective costs, in¯uences what central banks do. Since these
changes were made at a time of high employment and rapid money growth,
the experience suggests that, at times, fragility may enter lexicographically in
the central bank objective function. The size and direction of change in real
money balances may be useful measures of some of the costs of systemic
®nancial fragility.
3 Econometric evidence and interpretation
The three historical episodes discussed earlier suggest that an interest rate does
not fully represent the monetary transmission process. Changes in real money
balances appear at times to dominate changes in real interest rates as
indicators of the direction of change induced by monetary actions.
One interpretation of this evidence follows from the Haberler±Pigou±
Patinkin wealth effect. Falling prices raise private sector real money balances,
increasing real wealth. Some of the increase is spent on consumption. A
second interpretation views the change in real balances as a mixture of a pure
wealth effect and the changes in wealth induced by changes in the relative
prices of assets and output (Brunner and Meltzer 1963, 1968). On either
interpretation, we lose the simplicity of the two-equation, aggregate demand±
aggregate supply, class of models. If real balances affect aggregate demand, the
demand for money and perhaps other parts of the ®nancial structure become
relevant for the transmission process. The single interest rate is no longer
suf®cient to represent monetary policy and ®nancial markets.
Let me turn the issue around. The requirements of the standard model are
strong and are, as we have seen, easily falsi®ed empirically using historical
episodes. Analytically, the model is very demanding. Not only must the
wealth effect remain small and inconsequential for spending, but a single
interest rate must represent all relative price adjustment.
An appeal to rational expectations puts additional burdens on the model.
Market makers and participants must rationally anticipate how much output
prices will change to restore equilibrium at the natural rate. One data point, one
observation on the interest rate, is not suf®cient to determine whether the change
is permanent or temporary, a change in level or in growth rate, or an error that
will be reversed. Even if one believes that adjustment is rapid in markets for
actively traded assets, many assets are traded infrequently or not at all.
The transmission of monetary policy involves adjustment of prices of
existing hotels to newly produced hotels, existing plants and equipment to
newly constructed plants and equipment, new cars to used cars, new houses to
existing houses ± and, of course, money wages to prices.
I do not propose adding equations for each of these variables. That would
take us toward large-scale econometric models, a technology that has not
contributed greatly to our understanding of monetary policy. But there is also
considerable evidence that the expectations theory of the term structure or
interest arbitrage theories of the exchange rate are far from adequate. This
latter evidence suggests, again, that a single interest rate does not summarise
adequately responses on asset and output markets.
Recent analyses of the credit channel exploit differences in information to
develop an additional transmission channel for monetary actions (Bernanke
and Gertler, 1995). The credit channel operates parallel to the interest rate (or
monetary) channel, and the two interact. Empirical support for this channel
has not been persuasive; the hypothesised effect is probably swamped by the
endogenous response of bank loans to the monetary base and other
determinants of aggregate demand. The use of costly information to separate
parts of the transmission process is a step forward, however. It may ®nd a
better use distinguishing markets for bonds and real capital or distinguishing
different types of capital.
How well do real money balances capture the many channels of monetary
transmission? Koenig (1990) tested a two-stage model of changes in
consumption with changes in real money balances, real interest rates, income
and other variables as arguments of the consumption function. Consumption
includes only purchases of non-durables and services. The interest rate is
adjusted for in¯ation. All variables other than the interest rate and a dummy
variable are in logarithms. The dummy variable takes a value of unity in 1980
and after to represent ®nancial deregulation. All economic variables are used as
®rst differences.
Using data for the period from 1951:2 through 1986:1, Koenig found that
changes in real money balances had a signi®cant positive effect on changes in
consumption. To test his ®nding, Koenig introduced lagged values of changes in

durable goods, income, consumption, and stock prices sequentially as additional
explanatory variables. None was signi®cant at the 5 per cent level; in all cases the
change in real money balances remained positive and signi®cant.7 Furthermore,
he ®nds evidence of a relatively strong real balance effect. A 10 per cent change
in real balances, ceteris paribus, results in a 2.5±3 per cent change in spending on
consumption. This result is in addition to any effect of real balances on the real
interest rate and, thus, indirectly on consumption. Koenig found the latter effect
relatively small, however. He used a short-term interest rate.
In the three historical episodes discussed earlier, real balances rose in each
recession. In two of the recessions, the increase was more than 13.5 per cent
from peak to trough. In 1948±9, the increase was less than 1 per cent from peak
to trough but three±four times larger if measured one month before the
trough. If Koenig's estimates apply to these out of sample cycles, they imply a
3.5±4 per cent increase in consumption spending in the two earlier recessions.
This result is economically meaningful.
To test this hypothesis, I attempted ®rst to replicate Koenig's results. Then I
extended his sample period by 30 per cent, an addition of forty-two quarterly
observations. Most of the attempts at replication supported his hypothesis.
However, the change in the real interest rate was usually signi®cant and
positive, contrary to the hypothesis.
Koenig presents several different two-stage least square estimates. I was able
to replicate many of his ®ndings. Table 3.1 reports only a small part of the
results, but many are of interest principally because the main results are robust
to changes in the arguments included in the estimation. All results are two-
stage least squares estimates.
Table 3.1 Response of consumption to real balances and other variables, 1951:2±1986:1
Koenig (1990) Replication
Constant 0.0058* 0.0056* 0.0064* 0.0061*

(0.0005) (0.0006) (0.0004) (0.0007)
Dummy ±0.0048* ±0.0047* ±0.0056* ±0.0055*
(0.0012) (0.0012) (0.0011) (0.0012)
r 0.1948 0.1949 0.4490* 0.4458*
(0.1247) (0.1235) (0.1740) (0.1738)
ln m 0.2851* 0.2738* 0.3046* 0.2907*
(0.0588) (0.0624) (0.0851) (0.0910)
ln c(t ± 1) 0.0307 0.0258
(0.0754) (0.0778)
R2 0.37 0.37 0.37 0.37
DW 1.94 1.99 1.99 2.02
Notes:
Standard errors in parentheses.
*Signi®cant at the 5 per cent level.
Table 3.2 Response of consumption to real balances: extended sample, 1950:4±1995:4
m Real M1 m Monetary base
Constant 0.0064* 0.0060* 0.0060*

(0.0005) (0.0007) (0.0005)
Dummy ±0.0044* ±0.0042* ±0.0049*
(0.0009) (0.0010) (0.0010)
r 0.3212* 0.3090* 0.3105*
(0.1293) (0.1333) (0.1369)
ln m 0.2688* 0.2542* 0.3193*
(0.0583) (0.0590) (0.0925)
ln c(t ± 1) 0.0419
(0.0747)
R2 0.13 0.15 0.14
DW 1.91 1.99 1.98
Notes:
Standard errors in parentheses.
* Signi®cant at the 5 per cent level.
Table 3.2 shows some results for the extended sample. The ®rst two columns
use changes in real M1 balances, as in Koenig's work. The last two columns
replace M1 with the St Louis monetary base, the variable I used in the
historical data after 1935. Other estimates using lagged values of money, real
income, stock prices and other variables give similar results. The only change
that made a substantial difference was omission of the dummy variable used to
shift the intercept after 1980.
These ®ndings suggest, again, that the transmission of monetary impulses
involves more than is found in the standard class of models. Changes in real
money balances, operating as a wealth effect or as a proxy for changes in both
relative prices and real wealth, have positive and signi®cant effects on the
change in consumption.
4 Conclusion
Viewed one way, the point of this chapter is an old one, one that is well-known
as the Haberler±Pigou±Patinkin real wealth effect. The conclusion in that case
would be that the wealth effect is more relevant in a dynamic context than
current, standard analysis recognises.
An alternative interpretation is that the dynamic real wealth effect includes
more than the direct effect of changes in real money balances. If anticipated
returns and anticipated in¯ation affect prices of assets with low transaction
costs more rapidly than the prices of new production, many changes in
relative prices of assets to output are part of the transmission process. As home
prices change relative to foreign prices, the real exchange rate changes. This
change is supplemented by a change in the nominal exchange rate in a

¯uctuating rate regime.
The standard model includes a type of Phillips curve, so prices do not adjust
instantly to changes in money. Hence real balances change cyclically. The
consumption equation in the text suggests that the effects of monetary change
are not fully expressed by a change in a single short-term interest rate. A study
of forecast performance in Germany suggests that this ®nding is not limited to
the United States (Kirchga ssner and Savioz, 1998).
Rather than rely on a single regression equation, I have used data from
periods of de¯ation to show that changes in real interest rates cannot explain
some major episodes in monetary history. In 1920±1, nominal interest rates
were higher at the trough than at the preceding peak. Prices fell sharply, so real
interest rates were higher still. Postwar ®scal policy was contractive.
Why did the economy recover? Falling prices raised existing real wealth and
increased the gold ¯ow to the United States. Real money balances rose rapidly.
The same process worked to end the 1937±8 recession and, more modestly, the
1948±9 recession. Even in the 1929±33 depression, real balances played a role.
This time, the money stock fell in most periods by more than the price level, so
the real wealth effect did not sustain expansion.
Evidence from periods of de¯ation is useful because de¯ation raises both real
interest rates and real money balances. Positive changes in interest rates and
money have different implications: one points to expansion and an end to
recession, the other to continued, possibly deeper recession. In periods of
in¯ation, real interest rates are often low in recession, while real money
balances may fall. Monetary expansion reduces interest rates and raises real
balances. Both transmission processes are at work, so it is more dif®cult to
reliably separate their relative importance. Econometric work of the past
quarter-century testi®es to the dif®culty of distinguishing the two effects
when prices are rising.
Secular and permanent changes in real balances supplement the cyclical
changes that I have stressed. The introduction of the Euro is one such
permanent change. The breakdown of Bretton Woods, ®nancial intermedia-
tion and other long-term changes affected relative demands for real balances.
Such changes have persistent effects that, at times, dominate short-term
changes. Currency crises also are transmitted through changes in demand for
real balances as well as through interest rates and exchange rates. The current
standard model neglects these secular changes in the demand for real balances.
Central bankers choose to conduct their operations by changing an interest
rate, usually a short-term rate. This need not be harmful. The central bank
must adjust the interest rate enough to prevent de¯ation and in¯ation, having
regard to the cost of maintaining zero expected in¯ation. Whether making a
discretionary judgment or following a rule, history suggests that the bank will
avoid large, costly errors if it does not ignore the role of changes in nominal
money and in real balances when making its decisions.
Notes
* This chapter is a revised version of a dinner speech given by Allan Meltzer at the
conference.
1. My thanks to Randolph Stempski for his excellent assistance and to Bennett
McCallum for helpful discussion.
2. A similar but more complicated story is told about changes in money growth. Issues
about super neutrality are put aside.
3. I base my discussion on Svensson (1998). Other references could be used ± for
example, Taylor (1993, 1999).
4. A more complicated proof of this argument is in Brunner and Meltzer (1968).
5. For this period, the monetary base is high-powered money from Friedman and
Schwartz (1963). The price index is not seasonally adjusted.
6. Some may object that studies of Taylor's (1993) rule support use of the objective
function as currently relevant. Orphanides (1998) argues that this evidence is much
less compelling if we restrict the Federal Reserve to data available at decision time.
7. De®nition of variables is in Koenig (1990, pp. 422±4). We have used his de®nitions
wherever possible, but we are uncertain about his measurement of interest rates ±
end of quarter, quarterly average, etc. Interest rates are after tax rates, where tax rates
are marginal rates from Barro and Sahasakul (1983). We assumed constant marginal
tax rates after 1980.
References
Akerlof, G. A. Dickens W. T. and G. L. Perry (1996) `The Macroeconomics of Low
In¯ation', Brookings Paper on Economic Activity, 1, 1±76.
Balke, Nathan S. and R.J. Gordon (1996). Àppendix B: Historical Data' in R.J. Gordon
(ed.), The American Business Cycle: Continuity and Change. Chicago: University of
Chicago Press for the National Bureau of Economic Research, 789±842.
Barro, R. J. and C. Sahasakul (1983) `Measuring the Average Marginal Tax Rates from the
Individual Income Tax', Journal of Business, 56, 419±52.
Benhabib, J., S. Schmitt-Grohe and M. Uribe (1998) `The Perils of Taylor Rules', C. W.
Starr Center for Applied Economics, New York University, November, unpublished.
Monetary Policy Transmission', Journal of Economic Perspectives, 9, 27±48.
Brunner, K. and A. H. Meltzer (1963) `The Place of Financial Intermediaries in the
Transmission of Monetary Policy', American Economic Review, 53, 372±82.
ÐÐÐÐ (1968) `Liquidity Traps for Money, Bank Credit and Interest Rates', Journal of
Political Economy, 76, 1±37.
Friedman M. and A. Schwartz (1963) A Monetary History of the United States, 1867±1960,
Princeton: Princeton University Press for the National Bureau of Economic Research.
Ito, T. (1998) `Japan and the Asian Financial Crisis: The Role of Financial Supervision in
Restoring Growth', paper presented at a conference on the `The Japanese Financial
System', New York, Columbia University, 1±3 October.
Kirchga ssner, G. and M. Savioz (1998) `Monetary Policy and Forecasts for Real GDP
Growth: An Empirical Investigation for the Federal Republic of Germany', University
of St Gallen, unpublished.
Koenig, E. F. (1990) `Real Money Balances and the Timing of Consumption', Quarterly
Journal of Economics, May, 399±425.
Krugman, P. (1998) `Japan's Trap', Krugman web page, MIT, May.
Meltzer, A. H. (1995) `Monetary, Credit, and (other) Transmission Processes', Journal of
Orphanides, A. (1998) `Monetary Policy Evaluation with Noisy Information', Board of

Governors of the Federal Reserve System, Finance and Discussion Series, 1998±50,
unpublished.
Orphanides, A. and V. Wieland (1998) `Price Stability and Monetary Policy Effectiveness
when Nominal Interest Rates are Bounded at Zero', Board of Governors, Finance and
Discussion Series, 1998±35, unpublished.
Summers, L. (1991) `How Should Long-Term Monetary Policy Be Determined?', Journal
of Money, Credit and Banking, 23, 625±31.
Svensson, L. (1998) `Monetary Policy Issues for the ESCB', paper presented at the
Carnegie ± Rochester Conference, November.
Taylor, J. B. (1993) `Discretion versus Policy Rules in Practice', Carnegie ± Rochester
Conference Series on Public Policy, 39, 195±214.
ÐÐÐÐ (1999) Monetary Policy Rules, Chicago: University of Chicago Press.
4
Asymmetric Interest Rate Policy in

Europe: Causes and Consequences
Axel A. Weber
1 Introduction
The recent theoretical and empirical literature on monetary policy rules has
increasingly focused on short-term interest rates rather than monetary
aggregates for studying European monetary policy issues. There are several
reasons for this: ®rst, as in the United States, monetary aggregates in Europe
have displayed a less obvious link to real economic activity and in¯ation
during the 1980s and 1990s as opposed to the 1960s and 1970s. Second, many
central banks have de-emphasised the role of monetary aggregates and have
moved to operating procedures that focus more on interest rates (i.e. the Fed
funds target rate in the United States) or in¯ation rates (i.e. the in¯ation targets
in the United Kingdom, Canada, or New Zealand). Following the paper by
Taylor (1993) and more recent applications by Clarida and Gertler (1997),
Clarida, Gali and Gertler (1997, 1998), Gerlach and Smets (1998), Kuttner and
Posen (1998) and Rudebusch and Svensson (1998) there is now a growing
literature on so-called ìnterest rate smoothing' rules for Europe.1 These papers
use a simple policy reaction function in which interest rate adjustment
towards equilibrium depends on the deviations of in¯ation and output from
their respective target values. It is shown that such policy reaction functions ®t
the data quite well.
The present chapter takes a critical look at this new literature and discusses
two major problems. The ®rst of these problems is related to the interpretation
of such policy rules. It is argued that it is impossible to disentangle the causes
and consequences of policy actions in any policy reaction function, be it a
monetary policy rule or an interest rate policy rule, without imposing serious
prior restrictions. Unfortunately, such restrictions are typically dif®cult to
justify and to test. Think of the Clarida, Gali and Gertler (1998) paper, which
uses the level of short-term interest rates as the policy instrument and the
twelve-month ahead expected ( actual) in¯ation rate as well as the output
gap as the exogenous variables. If a restrictive monetary policy of increasing
interest rates in response to expected in¯ation is effective in easing
131
132 Axel A. Weber
in¯ationary pressure, then future actual in¯ation and hence in¯ation

expectations will decline as a consequence of the policy. Clarida, Gali and
Gertler (1998) realise this, since they state that `the expectations [of in¯ation]
on the right hand side [should] be based on all relevant information available
to policymakers ..., which includes the current interest rate itself, since
expected in¯ation and output will not be invariant to it'. They take account of
this endogeneity by using an instrumental variable2 technique. But single-
equation models can only indirectly capture such endogeneity, and the risk of
a serious bias of the estimated coef®cients remains. As an alternative we
propose to use the bivariate VAR model developed by King and Watson (1997)
in order to identify the in¯ationary causes and consequences of interest rate
smoothing policies more directly. In particular, we take a closer look at the
sensitivity of the estimated in¯ation response coef®cient () with respect to
the exogeneity assumption.3 This is of key importance since the magnitude of
the in¯ation response coef®cient provides an important yardstick for the
evaluation of such policy rules: if > 1, the target real rate adjusts to stabilise
in¯ation, whilst for < 1 it instead moves to accommodate changes in
in¯ation. The long-run Fisher effect is the critical threshold. As Clarida, Gali
and Gertler (1998) state, the central bank in this latter case raises the nominal
rate in response to an expected rise in in¯ation, but it does not do it
suf®ciently to keep the real rate from declining. Bernanke and Woodford
(1996) and Clarida, Gali and Gertler (1998) have shown that in such an
accommodative regime self-ful®lling bursts of in¯ation may be possible.
There is a second problem with interest rate smoothing rules which will be
brie¯y addressed here. We argue that it is impossible to discuss European interest
rate policy rules in isolation. Even if the short-run (or high frequency) cross-
country correlations in interest rates are small, there are clearly common long-
run factors behind European interest rate movements. In a dynamic bivariate
VAR model these long-run correlations are identi®able, and it will be possible to
distinguish between long-run and short-run correlations. It will be shown below
that the long-run cross-country correlations are high throughout the ERM period
and have increased over time owing to nominal convergence. This ®nding is
already obvious from Figures 4.1a±4.1d, which display the evolution of short-
term money market rates, discount rates, long-term government bond rates and
in¯ation rates for Germany and a number of selected European countries. In the
pre-1983 sample interest rates and in¯ation rates show little co-movement, but
during the 1990s there is an increasing synchronisation of European interest and
in¯ation rates owing to the convergence pressure from the Maastricht treaty.
Nominal convergence is almost complete in late 1998 shortly before the move to
EMU. Such cross-country co-movements of European interest and in¯ation rates
have largely been ignored when it comes to estimating policy reaction functions.
If the ERM was really an asymmetric exchange rate arrangement in which
Germany has played the role of the anchor country, then estimating country-
speci®c policy reaction functions is of interest only for Germany. The policy rules
Asymmetric Interest Rate Policy In Europe 133
(a) Short-term money market interest rates

20
18
France
16
14
a France
12
10
Belgium
8
6
4
Germany
Netherlands
2
0
1971 75 79 83 87 91 95 99
(b) Discount rates (b) Discount rates

16
14
12
b Belgium
10
8
6
Netherlands
4
Austria
Belgium
Germany
2
1971 75 79 83 87 91 95 99
Figure 4.1 Interest rates and in¯ation, 1972:1±1998:12

(a) Short-term money market interest rates
(b) Discount rates
134 Axel A. Weber
(c) Long-term government bond interest rates

24
Italy
20
c
France
16
Netherlands
12
Italy
France
8
Germany
Germany
4
1971 75 79 83 87 91 95 99
Inflationtrates
(d) Inflation rates
26
22
Italy
18
14
France
10
Netherlands
6
Italy
2
Germany
–2
1971 75 79 83 87 91 95 99
Figure 4.1 continued

(c) Long-term government bond interest rates
(d) In¯ation rates
for France, Italy and the remaining ERM countries have to be speci®ed differently
and this is why Clarida, Gali and Gertler (1998) include the German interest rate
as an important determinant of interest rates in the other ERM countries. But
provided that German interest rate movements can be well explained in terms of
a Bundesbank policy reaction function with respect to German in¯ation and
German output growth, these two factors will also explain some proportion of
interest rate movements in France. Figures 4.1a±4.1c show that German and
French interest rates are correlated, and Figures 4.1d±4.2a show that the same is
true for both in¯ation and output growth, but less so for money growth in Figure
4.2b. From the data's point of view it therefore seems to be unclear whether
interest rate smoothing rules in the ERM member countries like France are best
estimated on domestic or German data or any combination of the two. This is
why we aim at studying the degree of interest rate interaction between European
countries rather than looking at these countries in isolation.
The remainder of the chapter is organised as follows: Section 2 brie¯y
surveys the existing evidence about interest rate smoothing rules for Europe
while Section 3 deals with the ability of interest rate smoothing policies to
achieve the ®nal objective of monetary policy, which is low in¯ation. To get a
handle on this we use a bivariate VAR with in¯ation and interest rates.
Orphanides (1999) refers to this model as a simple prudent policy rule. We
study the short-run and long-run interaction of both variables in this prudent
rule under the very ¯exible identi®cation scheme proposed by King and
Watson (1997). Section 4 deals with the interaction of German and other
European interest rates, and we will examine the degree of long-run and short-
run interest rate co-movements. Section 5 concludes.
2 Empirical evidence about interest rates smoothing rules for

Europe
In this section we brie¯y review some of the existing evidence regarding

interest rate smoothing rules. It is important to note that such rules may be
viewed as a generalisation of the simple interest rate rule proposed by Taylor
(1993), which received widespread attention. In the typical Taylor rule the
central bank responds with the nominal interest rate it in a prescribed
manner to some known target deviations of in¯ation t � and output
yt � y :
it r t 0:5yt � y 0:5t � "t 4:1
Gathering all constant terms in (4.1) results in the typical Taylor rule in which
nominal interest rates it respond to output yt and in¯ation t :
it r � 0:5y 0:5yt 1:5t "t 4:2

|{z}
r~
136 Axel A. Weber
(a) Output (industrial production) growth in Germany and France

16
Germany
12
Germany
8
4
a
0
–4
France
–8
France
–12
–16
1971 75 79 83 87 91 95 99
(b) Money (M1) growth in Germany and France

22
18
14
b
10
6
2
–2
–6
1971 75 79 83 87 91 95 99
Figure 4.2 Output and money growth, 1972:1±1998:12

(a) Output (industrial production) growth in Germany and France
(b) Money (M1) growth in Germany and France
whereby the in¯ation response coef®cient has a numerical value of 1.5, and
hence real interest rates rise in response to in¯ation. Clarida, Gali and Gertler
(1997, 1998) and Rudebusch and Svensson (1998) introduce a partial
adjustment mechanism for interest rates owing to adjustment costs and
derive the dynamic interest rate smoothing rule:
it 1 � 1 � t 1 � yt it�1 "t 4:3
whereby 0 < < 1 holds for the partial adjustment coef®cient. The estimates
obtained by these authors are reproduced in Table 4.1. The key point about
Table 4.1 is that the long-run in¯ation response coef®cient is larger than unity
for Germany, Japan and the United States, but not for the United Kingdom,
France and Italy. In Clarida, Gali and Gertler (1998) the same result also applies
in many of the auxiliary regressions which include other potential policy
target variables. Rudebusch and Svensson (1998) use a modi®ed interest rate
smoothing rule in their optimal policy simulations, which are based on:
it 0:88t 0:30t�1 0:38t�2 0:13t�3

1:30yt 0:33yt�1 0:47it�1 0:06it�2 0:03it�3 "t 4:4
In contrast to the simple Taylor rule their optimal rule involves a long-run
in¯ation coef®cient of larger than 2, whilst the long-run output coef®cient in
the optimal rule is greater than 1.4 Kuttner and Posen (1998), on the other
hand, estimate the interest rate rule in ®rst differences. One of the rules they
estimate is the simple prudent rule:
it1 it1 � it c t � t�1 "t 4:5
Table 4.1 Selected coef®cient estimates of Clarida, Gali and Gertler (1998), baseline
regressions
Lagged AR (p)
Country Period In¯ation Output interest rate Constant model
Germany 79.03±93.12 1.31 (0.09) 0.25 (0.04) 0.91 (0.01) 3.14 (0.28) AR(1)
Japan 79.04±94.12 2.04 (0.19) 0.08 (0.03) 0.93 (0.01) 1.21 (0.44) AR(1)
USA 79.10±94.12 1.79 (0.18) 0.07 (0.06) 0.92 (0.03) 0.36 (0.85) AR(2)
USA 82.10±94.12 1.83 (0.45) 0.56 (0.16) 0.97 (0.03) ±0.1 (1.54) AR(2)
UK 79.06±90.10 0.98 (0.09) 0.19 (0.04) 0.92 (0.01) 5.76 (0.69) AR(1)
France 83.05±89.12 1.13 (0.07) 0.88 (0.10) 0.95 (0.01) 5.75 (0.28) AR(1)
Italy 81.06±89.12 0.90 (0.04) 0.22 (0.08) 0.95 (0.01) 7.14 (0.37) AR(1)
Coef®cient
range 0.90±2.04 0.08±0.88 0.92±0.97 ±0.1±7.14
Simple
Taylor rule 1.5 0.5
Source: Clarida, Gali and Gertler (1998).

138 Axel A. Weber
Table 4.2 Selected in¯ation coef®cient estimates of Kuttner and Posen (1998)
Country Pre-in¯ation In¯ation targeting Source in Kuttner

targeting period period and Posen (1998)
UK 1.38 (0.62) 2.01 (0.76) Table 7, Col. (a)

Canada 2.84 (1.35) ±0.67 (0.43) Table 4, Col. (a)
New Zealand 3.67 (0.30) 10.95 (1.41) Table 9, Col. (a)
Source: Kuttner and Posen (1998).
which corresponds closely to the type of interest rate policy rule we will
analyse below. The corresponding -estimates are summarised in Table 4.2.
Note that the in¯ation response estimates are typically larger than unity and
that in¯ation is assumed to be exogenous in Kuttner and Posen (1998). In the
analysis below we will see the potential impact of the possible endogeneity of
in¯ation on these coef®cient estimates.
3 Are interest rates and in¯ation linked in the long run?
Let us now turn to this endogeneity issue. Our key argument is that estimating
policy reaction functions is uninformative owing to the problem of
observational equivalence. Take the link between interest rates and in¯ation,
with the latter typically being considered as the monetary policy objective:
without detailed knowledge about the time series properties of the in¯ation
process, reduced form econometric methods are unable to discriminate
empirically between an interest rate policy reaction function (an anti-in¯ation
feedback rule) on the one side and the in¯ationary consequences of the
interest rate policy on the other side. However, as McCallum (1984) has
shown, such tests can be constructed using cross-equation restrictions in a
bivariate vector-autoregressive (VAR) approach.
Like much of the recent literature, we study a VAR representation of policy
and the economy, and we address an important criticism of VAR methods
concerning the possible non-robustness of identi®cation. To be precise, we
aim at relaxing the parameter restrictions required for identi®cation. In
particular, instead of focusing attention on the implications of a single set of
point estimates, we consider a sequence of identi®cation schemes satisfying
economically motivated inequality constraints. This estimation technique is
in the spirit of robustness checks exempli®ed by the studies of Blanchard
(1989), King and Watson (1997) and Bernanke and Mihov (1998). We limit
ourselves to the analysis of bivariate VARs, since the nature of this simple
structural model allows us to relate the long-run comovements of interest rates
and in¯ation to interpretable parameters describing the short-run dynamics of
both series in a bivariate setting. In particular, we specify the space of
admissible structural parameters and examine how our tested propositions

fare as the parameters vary over that space.
In what follows we will focus on the long-run link between in¯ation and
interest rates. We argue that this is what matters for a central bank's interest
rate policy strategies. Consider the long-run Fisher effect: although many
controversies surround the question of how, and to what extent monetary
policy can affect the economy in the short run, the long-run neutrality (LRN)
of money is widely accepted among policy-makers and monetary economists.
LRN means that nominal shocks do not have signi®cant effects on real
quantities such as output, employment and real interest rates in the long run.
It is further important to note that LRN implies the long-run homogeneity
(LRH) of nominal variables: nominal interest rates and in¯ation must change
proportionally in response to a nominal shock, which is the well known
prediction of the so-called Fisher effect.
Previous empirical evidence about the Fisher effect

Although LRN and LRH propositions appear to be commonly accepted (as
illustrated, for example, by their prominence in undergraduate macroeco-
nomics textbooks), formal empirical tests have yielded surprisingly mixed
results. One set of studies looks at the long-run Fisher effect from a cross-
country perspective. Using restricted least-squares regression techniques for
data averaged over long periods in a cross-section of countries, Lothian (1985)
tests the Fisher relation for twenty OECD countries and Duck (1988, 1993) for
a set of thirty-three countries (sixteen industrialised and seventeen developing
countries). While Duck (1993) ®nds cross-country evidence in favour of a
long-run Fisher effect, Lothian (1985) reports a less than proportional effect of
in¯ation on nominal interest rates, thereby rejecting the Fisher hypothesis.
Another set of studies, such as Lucas (1980), Mills (1982), Geweke (1982, 1986)
and Summers (1983), have attempted to test long-run economic relationships
by means of frequency-domain time series techniques. Summers (1983) uses
band-spectrum regression techniques to obtain low-frequency estimates of the
effects of in¯ation on real interest rates in a Fisher relation and rejects the
Fisher relation for post-war United States data. In a third class of papers,
Geweke (1986), Stock and Watson (1988), Fisher and Seater (1993), Weber
(1994) and King and Watson (1997) draw inference about the long-run Fisher
proposition based on explicit tests of coef®cient restrictions in bivariate
vector-autoregressive models, while a single equation autoregressive moving
average representation is estimated in Mishkin's (1984) analysis of the Fisher
effect. The evidence reported in Weber (1994) and King and Watson (1997)
rejects the existence of a Fisher effect for United States data, while the results
reported in Mishkin (1984) are in favour of a Fisher effect for the United States,
the United Kingdom and Canada, but not for Germany, France, the
Netherlands or Switzerland.
140 Axel A. Weber
The present chapter follows Geweke (1986), Stock and Watson (1988),
Fisher and Seater (1993), Weber (1994) and King and Watson (1997) and bases
inference about long-run economic propositions on explicit tests of coef®cient
restrictions in bivariate vector-autoregressive models.5 But in order to be able
to test for long-run neutrality (homogeneity) it has been shown that
meaningful tests can be constructed only if both monetary and real variables
satisfy certain non-stationarity conditions, which are spelled out in detail in
Fisher and Seater (1993) and King and Watson (1997). These studies
demonstrate that straightforward neutrality tests, such as imposing the
restriction that the coef®cients of current and lagged monetary impulses in
a regression on real economic variables sum to zero, can be conducted only if
the order of integration of both series is at least one and equal for both series.6
In addition, both series should not be co-integrated.7 Fisher and Seater (1993)
further show that much of the evidence in the older literature on long-run
neutrality or homogeneity violates these non-stationarity requirements, and
hence has to be disregarded.8 Before we present any structural estimates it is
therefore key to discuss brie¯y the unit root and cointegration properties of
the data.
The time series properties of interest rates and in¯ation

As mentioned above, our testing strategy critically depends on the relative
order of integration of the data. But these data properties also matter for the
approach of Clarida, Gali and Gertler (1998), who explicitly state that their
econometric approach relies on the assumption that within their short sample
both short-term interest rates and in¯ation are I(0).9 They ®nd that standard
Dickey±Fuller tests reject the null hypothesis that in¯ation is I(1) in G-3
economies in favour of the alternative of stationarity. For Germany they reject
that the short-term interest rate is I(1), while for the United States and Japan
there is less evidence against an I(1) process in interest rates. This would
preclude most of our tests. Thus, before presenting any of our results of the
Fisher hypothesis, it is important to discuss at some length the unit root
properties of the data.
The assessment in Tables 4.3 regarding in¯ation rates and nominal as well as
real interest rates is based on augmented Dickey±Fuller (ADF) `t-statistics', but
we also report Stock's (1991) 95 per cent con®dence intervals for the largest
unit root ().
Table 4.3 shows that nominal interest rates are non-stationary (I(1)) in
Germany, Italy, the Netherlands, Belgium, the United Kingdom and the
United States, and trendstationary (I(0)) in France, Denmark and Japan.
In¯ation rates, on the other hand, are non-stationary in all countries under
study. This is in contrast to the above assessment of Clarida, Gali and Gertler
(1998). But it is well known that non-stationarity tests have notoriously low
power in discriminating a unit root from an autoregressive coef®cient close to
unity, especially in the short samples we are studying. Nevertheless, our
®nding of unit roots in in¯ation and interest rates is consistent with previous
non-stationarity ®ndings in Gali (1992), Weber (1994) and King and Watson
(1997), who also ®nd indications of unit-roots in US in¯ation and interest data
during the postwar period and use both variables in ®rst difference form in
their VARs.
Table 4.3 Unit-root test statistics
ADF Drift, Stock's

Country Period Test Trend Stock's intervals Decision
(1) (2) (3) (4) (5) (6) (7)
(A) Nominal interest rates
Germany 79:03±98:11 ±2.43 T 0.98 ( ±, ± ) I(1)

France 79:03±98:11 ±3.66* T 0.93 ( ±, ± ) I(0)
Italy 79:03±98:11 ±3.19 T 0.95 ( ±, ± ) I(1)
Netherlands 79:03±98:11 ±1.51 ± 0.97 (0.96,1.02) I(1)
Belgium 79:03±98:11 ±2.86 T 0.90 ( ±, ± ) I(1)
Denmark 79:03±98:11 ±4.09** D,T 0.83 ( ±, ± ) I(0)
UK 79:03±98:11 ±2.49 T 0.92 ( ±, ± ) I(1)
USA 79:03±98:11 ±2.87 T 0.94 ( ±, ± ) I(1)
Japan 79:03±98:11 ±3.61* T 0.96 ( ±, ± ) I(0)
(B) In¯ation rates
Germany 79:03±98:11 ±1.70 ± 0.98 (0.95,1.02) I(1)

France 79:03±98:11 ±0.88 ± 1.00 (0.98,1.02) I(1)
Italy 79:03±98:11 ±1.03 ± 1.00 (0.97,1.02) I(1)
Netherlands 79:03±98:11 ±1.69 ± 0.98 (0.95,1.02) I(1)
Belgium 79:03±98:11 ±2.36 T 0.97 ( ±, ± ) I(1)
Denmark 79:03±98:11 ±2.75 T 0.95 ( ±, ± ) I(1)
UK 79:03±98:11 ±1.88 ± 0.98 (0.94,1.02) I(1)
USA 79:03±98:11 ±1.63 ± 0.99 (0.95,1.02) I(1)
Japan 79:03±98:11 ±1.82 ± 0.97 (0.94,1.02) I(1)
(C) Real interest rates
Germany 79:03±98:11 ±3.04* D 0.93 (0.88,1.00) I(0)

France 79:03±98:11 ±3.20* D 0.94 (0.87,1.00) I(0)
Italy 79:03±98:11 ±2.31 D 0.97 (0.92,1.01) I(1)
Netherlands 79:03±98:11 ±2.13 D 0.95 (0.93,1.01) I(1)
Belgium 79:03±98:11 ±2.69 D 0.90 (0.90,1.01) I(1)
Denmark 79:03±98:11 ±2.99* D 0.88 (0.88,1.00) I(0)
UK 79:03±98:11 ±2.03 D 0.94 (0.93,1.01) I(1)
USA 79:03±98:11 ±2.80 D 0.94 (0.90,1.01) I(1)
Japan 79:03±98:11 ±3.11 T 0.91 ( ±, ± ) I(1)
142 Axel A. Weber
Table 4.3 (continued)
ADF Drift, Stock's

Country Period Test Trend Stock's intervals Decision
(1) (2) (3) (4) (5) (6) (7)
(D) Nominal interest rate differentials (relative to Germany)
France 79:03±98:11 ±3.00 T 0.91 ( ±, ± ) I(1)

Italy 79:03±98:11 ±2.23 T 0.96 ( ±, ± ) I(1)
Netherlands 79:03±98:11 ±4.91** ± 0.82 ( ±,0.89) I(0)
Belgium 79:03±98:11 ±3.69* D,T 0.77 ( ±, ± ) I(0)
Denmark 79:03±98:11 ±4.91** D,T 0.72 ( ±, ± ) I(0)
UK 79:03±98:11 ±2.40 D 0.93 (0.91,1.01) I(1)
USA 79:03±98:11 ±1.95 ± 0.97 (0.94,1.01) I(1)
Japan 79:03±98:11 ±3.72* D,T 0.94 ( ±, ± ) I(1)
Notes:
Column 3 reports the augmented Dickey±Fuller tests for detrended data or demeaned data,
respectively. Their signi®cance levels are taken from Table 8.5.2. of Fuller (1976), p. 373.
A rejection of the null hypothesis of a unit root at the 1 per cent signi®cance level is marked with **,
at the 5 per cent level with *.
Stock's (1991) 95 per cent con®dence intervals for the largest unit root were calculated from the
ADF statistics using Stock's Tables A1 and A2 and the proceedure described in Appendix B of his
(1991) paper.
In addition to the con®dence belts for the estimated roots are displayed.
All ADF statistics are based on regressions including six lagged differences of the variable.
A second condition necessary for neutrality tests to be meaningful, as pointed

out by King and Watson (1992), is that nominal interest rates and in¯ation
must not be simultaneously non-stationary and cointegrated. The neutrality
hypothesis implied by the Fisher relation is that in¯ation has a zero long-run
effect on real interest rates and for this to be testable real interest rates must be
integrated of order one if in¯ation follows an I(1) process. This in turn requires
that nominal interest rates (rt ) and in¯ation (t ) must not be cointegrated,
since otherwise their linear combination is stationary. The unit root tests
reported for real interest rates in Table 4.3 suggest that except for Germany,
France and Denmark real interest rates also appear to be integrated of order
one from the ADF tests.
Based on this assessment we decided to carry out the neutrality tests
conditional on integrated processes, but we accept that in our short sample
some caution should be exercised in interpreting these results, since a
misspeci®cation might arise from the potential stationarity or cointegration in
the data for Germany, France and Denmark. However, our results also show
that a spurious regression problem may apply to the estimates of interest rate
smoothing rules speci®ed in levels of interest rates, in¯ation and output.10
An econometric framework for studying the long run link between

interest rates and in¯ation
McCallum (1984) points out that valid tests of long-run neutrality may be
constructed only by using cross-equation restrictions in a bivariate approach.
The present chapter follows King and Watson (1997) and bases inference
about long-run economic propositions on explicit tests of coef®cient
restrictions in bivariate vector-autoregressive models. In order to brie¯y
illustrate their approach, consider the hypothesis that a permanent change in
the in¯ation rate (pt t ) has no long-run consequences for the level of real
interest rates (rt ). Since real interest rates are typically measured as the
difference between nominal interest rates and in¯ation rates, this neutrality
proposition implies the homogeneity restriction that nominal interest rates move
one-to-one with in¯ation in the long run.
If we allow in¯ation to affect nominal interest rates (i.e. owing to the Fisher
effect) and at the same time assume that in¯ation reacts endogenously to
movements in interest rates rates (i.e. as a result of the effects of an interest rate
policy) we obtain a bivariate vector-autoregressive (VAR) model in both
in¯ation and interest rates:
X
p
j
X
p
t i it i it�j j t�j "m
t 4:6
j1 j1
X
p
j
X
p
j
it i t ii it�j i t�j "t 4:7
j1 j1
where i and i represent the contemporaneous effect of interest rate policy
on in¯ation and the contemporaneous response of interest rate policy to
in¯ation, respectively. A more convenient representation of this bivariate VAR
system is:
Lt i Lit "m

t 4:8
ii Lit i Lt "t 4:9
whereby
Pp j Pp j Pp j
ii L 1 � j1 ii Lj ; i L i j
j1 i L ; L 1 �
j
j1 L as well
Pp j j
as i L i j1 i L applies. In stacked form this may be re-written as:
LXt "t 4:10

Pp
where L j0 j Lj with
m
t " 1 �i
Xt ; "t t ; 0 ;
it "t �i 1
144 Axel A. Weber
" #
j j
i
and �
j
j j ; j 1; 2; :::; p
i ii
In the above notation the long-run multipliers are i and i; where
i i 1=ii 1 measures the long-run response of nominal interest rates
to a one unit permanent increase in in¯ation, while i i 1= 1
measures the long-run response of in¯ation to a permanent unit increase in
interest rates. The long-run Fisher effect implies the restriction i 1.
As noted by King and Watson (1997), (4.10) is econometrically unidenti®ed
and the restrictions implied by the Fisher effect are no longer testable when

in¯ation is endogenous. Thus, even if the hypothesis that "m t and "t are
uncorrelated is maintained, one additional restriction is required in order to
identify the linear simultaneous equation model. Common identifying
assumptions are that in¯ation is exogenous ( i 0) or predetermined
i 0. Alternatively, the long-run Fisher effect with i 1 may be imposed
in order to identify the system and estimate the remaining parameters. In
principle it is possible to identify the above model by specifying a value of any
one of the four parameters i ; i ; i or i and then ®nding the implied
estimates for the other three.
Empirical evidence regarding the long-run link between interest rates

and in¯ation
Table 4.4 reports the evidence for the ERM period. As compared to Clarida,
Gali and Gertler (1998) we analyse the slightly longer sample 1979:03 to
1998:10 and consider a broader set of European countries. The estimated
short-run correlation (column (5)) between nominal interest rates and
in¯ation for Germany is zero, and the long-run correlation (column (8)) of
0.79 is much higher but still substantially smaller than one. For France the
short-run correlations are somewhat higher (0.15) and the long-run
correlations somewhat lower (0.62) than for Germany, and the same holds
for Italy, the Netherlands, Belgium, and Denmark. The United Kingdom,
and Japan have higher long-run correlations than Germany, whilst the
United States displays one of the lowest long-run correlations. Imposing a
long-run Fisher effect i 1 furthermore results in some interesting
estimates of the remaining coef®cients (columns (12)±(14)). For example,
under a long-run Fisher effect in¯ation is found to be exogenous in the long
run in all countries. For Germany, France, Italy, the Netherlands, the
United Kingdom, the United States and Japan a signi®cant negative short-
run impact of domestic interest rate policy on in¯ation is found, but not for
Belgium and Denmark. For the small ERM countries this result is not too
surprising, given the high degree of openness and the importance of non
domestically determined traded goods prices in their consumer price
indices.
Table 4.4 Long-run link between interest rate policy and in¯ation in European and selected G7-countries
VAR Structural model i 1 in 95 per cent Estimates imposing i 1

estimates estimates con®dence interval
Country Period 2i 2 cori 2i 2 cori i i i i i i
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
Germany 79.03±98.10 0.31 0.30 0.01 0.49 0.75 0.79 ±1.8,0 0,1.5 ±4 0.18 (0.21) 0.48 (0.26) ±0.50 (0.28)
France 79.03±98.10 0.27 0.62 0.15 0.47 0.77 0.62 ±0.6,0 0.44 ±2.5 0.01 (0.19) 1.24 (0.51) ±0.18 (0.12)
Italy 79.03±98.10 0.31 0.55 0.04 0.71 0.73 0.66 ±0.04 0.26 0.16 ±0.86 (1.02) 1.13 (0.51) ±0.35 (0.18)
Netherlands 79.03±98.10 0.27 0.61 ±0.01 0.40 0.55 0.48 ±0.2 0.86 0.17 ±0.29 (0.32) 2.15 (0.86) ±0.43 (0.18)
Belgium 79.03±98.10 0.32 0.95 0.08 0.39 0.63 0.47 ±0.3,0.8 ±0.29 0.2 ±0.14 (0.21) 0.80 (0.55) ±0.06 (0.07)
Denmark 79.03±98.10 0.45 1.37 0.10 0.49 0.87 0.50 ±0.1,0.2 0.85 n.a ±0.05 (0.15) ±0.11 (0.48) 0.05 (0.06)
UK 79.03±98.10 0.49 1.08 0.17 1.14 0.86 0.86 ±0.2 1.16 n.a. 4.26 (4.02) 2.27 (0.72) ±0.48 (0.22)
USA 79.03±98.10 0.27 0.67 0.16 0.55 0.68 0.49 ±1.1,0 0.84 0.1 ±0.43 (0.40) 1.89 (0.63) ±0.27 (0.14)
Japan 79.03±98.10 0.44 0.29 ±0.01 0.54 0.70 0.88 0.02 0 ± 0.25 (0.22) 0.28 (0.16) ±0.67 (0.38)
Notes:
All results for the second moments are based on VARs with six lags.
2i denotes the variance estimate for variable i, corij indicates the correlation between variables i and j.
Variances and correlations are calculated for the residuals of the unrestricted VARs and the shocks implied by the long-run covariance matrix of the estimated
VAR (the spectral density matrix of the variables at frequency zero).
The coef®cient ranges in columns (9)±(11) are those for which the long-run homogeneity proposition cannot be rejected at the 95 per cent level.
The point estimates of the coef®cients and their 95 per cent con®dence regions (2 standard errors in parentheses) implied by long-run homogeneity are
reported in columns (12)±(14).
145
146 Axel A. Weber
Another interesting ®nding of Table 4.4 concerns the link between the long-
run and short-run in¯ation response coef®cients. Both the Taylor rule and the
interest rate smoothing rules of Clarida, Gali and Gertler (1998) have a built-in
long-run Fisher effect as the neutral policy stance under which the central
bank's interest rate policy neither actively stabilises nor accommodates
in¯ation. The Fisher effect is therefore useful as a benchmark case. By
imposing the Fisher effect ( i 1), we obtain estimates of the short-run
in¯ation response coef®cient (i ) which are signi®cantly larger than one for
the Netherlands, the United Kingdom and the United States, while for
Germany, France, Italy, Belgium and Japan they are signi®cantly greater than
zero but not larger than one. For the latter countries a less than proportional
interest rate adjustment to in¯ation appears to be consistent with a long-run
neutral policy stance, whilst for the Netherlands, the United Kingdom and the
United States the data point towards some degree of short-run interest rate
smoothing policies.
Since the short-run in¯ation response coef®cient for Germany is signi®-
cantly smaller than unity, does this imply that the Bundesbank has
accommodated in¯ation? The answer is no. Consider Figure 4.3, which
displays the coef®cient sensitivity analysis graphically. Economically mean-
ingful policy reaction functions clearly imply i 1. Panel b shows that for
Germany even moderately positive in¯ation response coef®cients
(0 < i < 1:5) are compatible with a long-run Fisher effect ( i 1). According
to Panel c the data also indicate that German in¯ation is in the long run
exogenous with respect to interest rates across a wide range of possible
identi®cation schemes, whilst from Panel a it is obvious that i 1 implies a
strongly negative immediate impact of interest rates changes on in¯ation (i ).
On the other hand, if we assume in¯ation to be predetermined (i 0), this is
consistent with i 1 only for an initial policy response to in¯ation which is
close to zero (�0:1 <i < 0:1).11 In general, Panel d of Figure 4.3 shows that
there exists a `tradeoff' between the coef®cients i ; and i . This allows the
reader to carry out the following `speci®cation test' concerning the long-run
Fisher hypothesis: if the reader believes that the true value of the pair (i ; i )
lies outside the 95 per cent con®dence region, then the model with the long-
run Fisher effect imposed is rejected by the data at the 5 per cent level. We
believe that monetary policy has long and variable outside lags, and therefore
in¯ation will initially only respond moderately to interest rate changes
(i 0). We furthermore believe that money is neutral in the long-run and
that a long-run Fisher effect exists. This together implies that monetary
policy of the Bundesbank is unlikely to be adequately described by an interest
rate smoothing rule, since a long-run unit effect of in¯ation on interest rates
for Germany is compatible with no short-run policy reaction of interest rates
to in¯ation at all. The same prior beliefs would lead us to conclude from
Figure 4.4 that for the United States only a much larger immediate response
of interest rates to in¯ation would be consistent with a long-run neutral
Figure 4.3 The link between interest rates policy and in¯ation, Germany, 1979:3±
1998:10
monetary policy stance. This empirical result may simply be a re¯ection of the
fact that the United States actually conducted its monetary policy through an
interest rate smoothing rule, whilst the Bundesbank did not.
For the remaining countries the results in Table 4.4 may be summarised as
follows: Overall, our results are quite similar for all countries, and Figure 4.5
demonstrates this for the con®dence ellipses under the null hypothesis of a
long-run Fisher effect. More speci®cally, our ®ndings for most ERM countries
resemble the results obtained for Germany, while the United Kingdom is more
like the United States. With the exception of the United States, the closest
long-run correlations between in¯ation and interest rates are typically found
for the large economies (Germany, United Kingdom, Japan), while smaller
economies display a lower long-run correlation. This may indicate that during
the ERM German interest rates rather than domestic in¯ation may have been
the dominant driving force behind domestic interest rates in many countries.
We will examine this proposition in more detail below.
148 Axel A. Weber
Figure 4.4 The link between interest rates policy and in¯ation,
United States, 1979:3±1998:10
4 Are European interest rates linked in the long run?
Since January 1999 European interest rate policy was placed under the authority
of the European Central Bank (ECB) and of®cial interest rates in EMU member
countries are, by de®nition, moving in a perfectly symmetric fashion. Prior to the
creation of the monetary union this was not the case, but owing to nominal
convergence there has been an increasing degree of co-movements between
European rates. In this context it has frequently been argued that Germany has
played the role of the anchor country in the ERM, and that interest rate changes
originating in Germany would ultimately cause symmetric changes in the interest
rates of the remaining ERM countries. To examine this proposition, we ®rst
analyse the data properties and then apply our estimation technique.
The time series properties of interest rate differentials

As already mentioned above, our testing strategy depends on the relative order
of integration of the data, which need to be I(1) and not cointegrated. The time
series properties of nominal interest rates were discussed on p. 140, and it was
shown that nominal interest rates are non-stationary, I(1), in Germany, Italy,
the Netherlands, Belgium, the United Kingdom and the United States but
trendstationary, I(0), in France, Denmark and Japan. For the latter countries
95% confidence ellipse when γyi,π=1

5
UK Netherlands
4
3
USA
Italy
2
Germany
λi,π
1
0
Belgium
–1
Denmark
–2
–1.2 –0.8 –0.4 –0.0 0.4
λπ,i
Figure 4.5 The estimated tradeoff between policy response and the policy effects for
interest rate policy and in¯ation, 1979:03±1998:10
Note: The coef®cients ;i and i; represent the contemporaneous effect of interest rate
policy on in¯ation and the contemporaneous in¯ation response of interest rate policy
respectively.
The con®dence epilses indicate the parameter space for the pair ;i ; i; over which
the data do not reject the long-run Fisher-effect at the 5 signi®cance level.
the tests below have to be interpreted with caution. But in addition to non-
stationarity, the absence of cointegration and hence the non-stationarity of

interest rate differentials vis-a-vis Germany is required. Based on augmented
Dickey±Fuller tests we cannot reject the non-stationarity of interest rate
differentials for France, Italy, the United Kingdom and the United States, while
trend-stationary processes are found for Belgium, Denmark and Japan and a
mean-stationary process is identi®ed for the Netherlands. We will discuss our
®ndings in detail only for the ®rst set of countries.
A VAR model of international interest rate linkages

Formally, the long-run cross-country proportionality of interest rates may be
studied by relating interest rate movements abroad (it ) to changes in the
German interest rates (it ):
it i i it "t 4:11
with a coef®cient ii 1 implying the perfect immediate transmission of

German interest rate innovations. But typically estimation of (4.11) is
150 Axel A. Weber
complicated by two facts. First, the interest rate policy may be transmitted
with lags, and in practice it therefore makes sense to distinguish between the
immediate and the long-run transmission of interest rate policy by using the
dynamic speci®cation:
X
p
j
X
p
j
it i i it i i it�j i i it�j "t 4:12
j1 j1
which allows the concept of long-run versus short-run policy transmission to

be implemented empirically. A second important complication in the
estimation of (4.11) or (4.12) arises from the fact that the German interest
rate it cannot safely be considered as an exogenous variable. For example, if
the German central bank is concerned about exchange rate movements and
therefore responds to interest rate movements abroad, German rates would
change in line with interest rates abroad. This endogeneity problem may
explicitly be taken into account by estimating the simultaneous two equation
system:
X
p
j
X
p
j
it i i it i i it�j i i it�j "m
t 4:13
j1 j1
X
p
j
X
p
j
it i i it ii it�j i i it�j "t 4:14
j1 j1
using the 2SLS instrumental variable techniques developed in King and

Watson (1997). This approach has two advantages over the previous literature:
®rst, the estimates explicitly allow German interest rate movements to be both
predetermined (i i 0) and/or exogenous in the long-run [ i i i i 1=i i 1
Pp j Pp
0 with lag polynomials i i L i i j1 i i L j and i i L 1 ± j1 ii i Lj ]
without necessarily imposing these restrictions onto the data. Second,
equations (4.13) and (4.14) provide a natural framework within which the
concept of immediate interest rate policy transmission (ii 1) and long-run
Pp i j
transmission [ ii ii 1=ii 1 1 with ii L ii j1 ii L and ii L
Pp j j
1 � j1 ii L ] can be formalised and tested empirically.
As noted above, the bivariate VAR system in (4.13) and (4.14) is
econometrically unidenti®ed. In the previous literature various identifying
restrictions are to be found: it is common practice in the older literature to
invoke the timing assumption and restrict German interest rate policy signals
to be long-run exogenous, so that i i 0 holds. Another approach is to
assume that the model is recursive, so that German interest rates are
predetermined i i 0. Alternatively we could impose ii 0. As before, we
will identify the model by pre-specifying a value for any one of the four
parameters ii ; i i ; i i or ii and then ®nding the implied estimates for the
other three parameters.
Empirical evidence about long-run international interest rate linkages

Before discussing the results it is important to note that equations like (4.13)
have previously been estimated in the context of the EMS asymmetry
literature by applying Granger-causality tests to European interest rates. Such
Granger-causality tests have been conducted by Cohen and Wyplosz (1989),
DeGrauwe (1989), Fratianni and von Hagen (1990a, 1990b), and Weber (1990)
in order to demonstrate the asymmetrical functioning of the EMS as a de facto
`DM zone'. However, the evidence from this literature has revealed a
surprisingly symmetrical working of the EMS, with causality typically running
both ways. Wyplosz (1991) has argued that this ®nding is hardly surprising,
given that the strategic (game-theoretic) policy-making literature suggests that
it is sub-optimal for policy makers to follow a predetermined or exogenous
policy without taking the other central bank's past or current policy reactions
into account. Equation (4.14) captures precisely this fact. The two-equation
system estimated here may thus be viewed as a more consistent approach to
testing for asymmetric interest rate policies if policy makers are known to
behave strategically and to care about each other's policy settings.
Some non-formal evidence about the high degree of European interest rate
policy interdependence was already visible in Figure 4.1. Discount rate policies
appear to be relatively closely co-ordinated between Germany, the Nether-
lands ± and, to a slightly lesser extent Belgium. The same three countries
display close co-movements of short-term money market rates, long-term
government bond rates and in¯ation rates. Figure 4.1 also showed that there
were short-run destabilising effects of the speculative attacks between
September 1992 and August 1993 on European money market interest rates
± in particular, for Italy, Ireland, Denmark and France. Overall, the impression
from Figure 4.1 was that interest rate co-movements within Europe were quite
pronounced.
In order to obtain formal evidence on the degree of asymmetry in interest
rate policies in Europe we have estimated the simultaneous two-equation
system (4.13) and (4.14). Table 4.5 reports the empirical results.
All the long-run correlations (column (8)) of interest rate innovations are
positive and fairly high for France (0.81) and the Netherlands (0.75), while
relatively low correlations are found for Italy (0.46) and the United Kingdom
(0.31). This is caused by the speculative attacks of September 1992, because the
exit of both countries from the ERM allowed their interest rates to move more
freely and more independently from the rest of Europe. The highest short-run
correlation is found for the Dutch±German interest rate, and thus points
towards a high degree of short-run policy credibility.
Columns (9)±(14) show our results when we impose a perfect long-run co-
movement of interest rates, which theoretically should be the case under
152
Table 4.5 Long-run cross-country in¯uence of German interest rates policy on interest rates in European and selected G7-countries
VAR Structural model ii 1 in 95 per cent Estimates imposing ii 1

estimates estimates con®dence interval
Country Period 2i 2i corii i2 2i corii ii ii ii ii ii ii
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
France 79.03±98.10 0.31 0.59 0.18 0.58 0.87 0.81 ±0.2,0.2 ±0.5,0.9 ±0.3 0.21 (0.14) 0.20 (0.31) 0.04 (0.09)
Italy 79.03±98.10 0.31 0.55 0.12 0.55 0.79 0.46 ±0.4,0.1 0,1.3 0.1 ±0.29 (0.24) 0.56 (0.29) ±0.11 (0.11)
Netherlands 79.03±98.10 0.30 0.56 0.35 0.52 0.56 0.75 ±0.2,0.1 0.1,1.2 0.2 ±0.48 (0.53) 0.71 (0.25) ±0.02 (0.09)
Belgium 79.03±98.10 0.31 0.91 ±0.01 0.53 0.64 0.59 ±0.3,0.0 0.2,2.8 0.1 ±0.39 (0.33) 1.12 (0.54) ±0.13 (0.06)
Denmark 79.03±98.10 0.31 1.44 0.13 0.55 1.33 0.53 ±0.1,0.1 ±1.0,1.8 ± 0.1 0.06 (0.09) 0.53 (0.73) 0.01 (0.04)
UK 79.03±98.10 0.31 1.15 0.11 0.54 0.83 0.31 ±0.3,0.0 0.9,3.6 0.0 ±0.27 (0.20) 1.91 (0.64) ±0.12 (0.06)
USA 79.03±98.10 0.29 0.70 0.04 0.57 0.87 0.54 ±0.8,±0.1 0.8,4.0 0.2 ±0.12 (0.20) 1.82 (0.64) ±0.30 (0.12)
Japan 79.03±98.10 0.31 0.31 0.20 0.55 0.71 0.37 ±0.2 0.4 0.6 ±0.44 (0.36) 0.80 (0.23) ±0.74 (0.35)
Notes:
All results for the second moments are based on VARs with three lags.
2i denotes the variance estimate for variable i, corij indicates the correlation between variables i and j.
Variances and correlations are calculated for the residuals of the unrestricted VARs and the shocks implied by the long-run covariance matrix of the estimated
VAR (the spectral density matrix of the variables at frequency zero).
The coef®cient ranges in columns (9)±(11) are those for which the long-run homogeneity proposition cannot be rejected at the 95 per cent level. The point
estimates of the coef®cients and their 95 per cent con®dence regions (2 standard errors in parentheses) implied by long-run homogeneity are reported in
columns (12)±(14).
EMU. Let me focus on the ERM countries here and disregard the free ¯oaters,
because for them this is an inadequate restriction.12 But even among the ERM
countries this restriction is inconsistent with a predetermined or a long-run
exogenous German interest rates policy. Except for France the estimated
coef®cients in columns (12) and (14) are predominantly negative and
frequently signi®cant. For the data not to reject the imposed restriction, the
Bundesbank would have had to increase German interest rates in response to
falling interest rates in the rest of Europe. During German uni®cation this was
in part the case, but the restrictive German monetary policy was motivated
exclusively by domestic objectives related to the tax versus credit ®nancing of
uni®cation and not in response to European developments. Furthermore, the
above restriction would imply a relatively high immediate impact of German
interest rate policy on interest rates abroad, ranging from 0.2 for France to 1.12
for Belgium. Except for France this by far exceeds the short-run correlations in
the data and we therefore reject this restriction.
When comparing Tables 4.4 and 4.5, another interesting result becomes
obvious. For the smaller ERM countries, such as the Netherlands, Belgium or
Denmark, the long-run correlations of domestic interest rates with German
interest rates are actually higher than the corresponding correlations with
domestic in¯ation. In this case the obvious interest rate policy rule appears to
be that these countries have predominantly followed interest rate movements
in Germany. But our estimates in Table 4.5 show that whilst interest rates
appear to be interconnected, there is no simple long-run homogeneous
pattern of co-movements during the ERM period.
5 Summary and conclusions
In this chapter we have taken a closer look at interest rate behaviour in

industrialised countries with special emphasis on Europe. We have argued that
it is inherently dif®cult to interpret interest rate movements simply as
re¯ecting policy actions in response to key economic variables such as
in¯ation or output growth. Rather than following the standard approach of
the current literature and estimating interest rate smoothing policy rules in
order to discuss their degree of asymmetry, we have analysed the joint
behaviour of interest rates and in¯ation under a variety of identifying
restrictions. For simplicity and clarity of the argument we have chosen a
bivariate setting with interest rates and in¯ation only. Orphanides (1999)
refers to this as a `simple prudent policy rule'.
In our analysis we inquire whether or not the data are consistent with a
long-run Fisher effect under endogenous interest rate policy choices. In this
case it is obvious that no simple interpretation of the joint behaviour of
interest rates and in¯ation is possible owing to the problem of observational
equivalence. This is a well known and old problem of single-equation
econometric policy evaluation, dating back to the early studies of Lucas (1972,
154 Axel A. Weber
1976) Sargent (1973) and McCallum (1984). What is surprising is the

consistency and persistency with which this problem is being ignored in
empirical work. The main reason for this is that no easily interpretable
econometric results follow from the type of sensitivity check which we have
performed in this chapter with respect to different identi®cation schemes. We
®nd that it is dif®cult to disentangle the causes and consequences of interest
rate policy even in our simple prudent policy rule which only incorporates two
variables. But imposing certain long-run neutrality characteristics (i.e. the
Fisher effect) helps us to distinguish between the causes and the consequences
of interest rate policy within a bounded set of parameter values.
Our ®ndings and conclusions are, of course, subject to some reservations:
®rst, to analyse `typical' policy rules our bivariate prudent policy rule would
have to be extended to a multivariate setting to avoid the danger of a serious
misspeci®cation. Second, the short sample of monthly observations for the
ERM period conveys only limited `long-run' information, and this seriously
undermines the power of both the unit root and the long-run neutrality tests
conducted in this chapter. But moving to longer samples involves the risk of
mixing different policy regimes. Both are serious drawbacks, and there is no
obvious solution. Third, the application of the neutrality tests developed by
King and Watson (1997) critically depends on the order of integration in the
data, which is subject to a large degree of uncertainty owing to the low power
of unit root tests. It is found that the sub-set of time series which possess the
postulated unit-root properties may be small, so that in many cases long-run
neutrality or homogeneity tests are applicable only to a limited extent.
The present chapter also raises another critical point concerning the
standard literature on interest rate smoothing policies, which has to do with
the degree of interdependence of interest rate policies across Europe. Again, it
is well known that interest rates are linked through the uncovered interest rate
parity (UIP) condition and it is hard so see how under an exchange rate
pegging arrangement like the ERM independent national interest rate policies
can be conducted. This is different under a free ¯oat, and our estimates
con®rm this. But since European interest rates are shown to move closely
together, estimating independent interest rate smoothing rules for various
ERM countries (such as Germany, France and Italy) in Clarida, Gali and Gertler
(1997) is problematic. For these countries movements in German rates need to
be accounted for when estimating interest rate policy reaction functions. For
the smaller ERM countries, such as the Netherlands, Belgium or Denmark the
long-run correlations of domestic interest rates with German interest rates are
actually higher than the corresponding correlations with domestic in¯ation.
But in spite of all the above mentioned limitations, we think that the King
and Watson (1997) methodology provides a powerful tool for analysing the
long-run predictions of economic theories within a fairly general framework.
In future research, we plan to extend our model to a three-variable VAR
including output, in¯ation and interest rates. However, as pointed out by King
and Watson (1997), the identi®cation problem becomes much more dif®cult
in this case, since the number of identifying parameter restrictions increases
with the square of the variables in the model. This may pose computational
problems, given that each identifying restriction is iterated hundreds of times
over a wide range of plausible values.
Appendix Time series and data sources
The chapter uses monthly data. The time series and data sources used were:
. Interest rates: Money market rates, International Monetary Fund (IMF),
International Financial Statistics, various issues, line 60b:
± Discount rates, IMF, International Financial Statistics, various issues, line 60.
± Treasury bill rates, IMF, International Financial Statistics, various issues,
line 60c.
± Government bond rates, IMF, International Financial Statistics, various
issues, line 61.
. Consumer prices: IMF, International Financial Statistics, various issues, line
64.
. Money stock (M1): IMF, International Financial Statistics, various issues, line
34.
. Output, industrial production: IMF, International Financial Statistics, various
issues, (line 66b)
Notes
1. Gerlach and Smets (1998) apply such a model to aggregate data constructed for the
Euro area, whilst the other studies focus on various countries separately.
2. Clarida, Gali and Gertler (1998) try to control for the endogeneity of expected
in¯ation by using past interest and in¯ation rates as well as other variables (output,
commodity prices and real exchange rates) as instruments in a two-step nonlinear
two-stage least squares estimator. For details see n. 11 of their paper.
3. Obviously, our choice to employ a bivariate VAR instead of a multivariate VAR
approach involves the risk of a potential misspeci®cation. However, the alternative
approach also suffers from a potential spurious or unbalanced regression problem
since output is typically not found to be a trendstationary process, as proposed by
Clarida, Gali and Gertler (1998). We ®nd that using output growth instead results
in at the best only marginally signi®cant coef®cient estimates. Thus, under a more
appropriate speci®cation the potential misspeci®cation problem largely reduces. In
our view this justi®es the use of a bivariate VAR, but we agree that extending our
model to a multivariate VAR is the obvious next step for future research.
4. See Rudebusch and Svensson (1998), p. 25.
5. Long-run neutrality here implies a zero-restriction on the sum of coef®cients of the
contemporaneous and lagged in¯ation variables in a regression on real interest
rates. Tests of short-run neutrality, such as those conducted in Sims (1972), on the
other hand, impose zero-restrictions on the individual coef®cients of the monetary
impulses. Long-run neutrality is thus a weaker test, and short-run non-neutralities
may well be compatible with long-run neutrality.
156 Axel A. Weber
6. In the present context unit root tests are important since they indicate those cases
for which the neutrality results may be a statistical artefact. This has been shown to
be the case if the two variables under study do not have the same relative order of
integration. There is no implication whatsoever with respect to the absolute order
of integration in the data, given that it is at least one. The absolute order of
integration does determine the appropriate minimum degree of differencing of the
data before conducting neutrality tests, but overdifferencing does not affect the
neutrality propositions. See Fisher and Seater (1993) for details.
7. If it and t are non-stationary and cointegrated, then a ®nite VAR in ®rst differences
does not exist because there would be a unit root in a moving average polynomial,
which is not invertable. Fisher and Seater (1993) brie¯y discuss cointegration in the
context of long-run neutrality restrictions in an Appendix. They show that
cointegration per se does not affect the long-run neutrality restrictions derived.
However, cointegration typically is suf®cient for rejecting long-run neutrality.
8. Long-run neutrality test results will be presented even if unit root tests suggest that
they may be uninformative. The key point is that the unit root tests may indicate in
which case the neutrality results are a statistical artifact.
9. See Clarida, Gali and Gertler (1998), n. 9.
10. Similar unit root tests for output (not reported here) show that in none of the
countries considered here is output (industrial production) driven by a trend-
stationary process, as is assumed in Clarida, Gali and Gertler (1998) for their
measure of the output gap. In line with much of the literature we ®nd output to be
an I(1) process.
11. Note that the corresponding estimates of i in Clarida, Gali and Gertler (1998) for
Germany are also very low (i 0:118) owing to the large AR(1) coef®cient in the
interest rate equation.
12. For the free-¯oating countries the data are clearly inconsistent with this restriction:
in most cases the contemporaneous effect of German interest rate changes on
foreign interest rates would be relatively high, taking minimum values of 0.4 for
Japan to 0.8 for the United States and 0.9 for the United Kingdom. This by far
exceeds the short-run correlations in the data. German interest rates would also be
endogenous and react strongly to interest rate settings abroad.
References
Bernanke, B. and I. Mihov (1998), `The Liquidity Effect and Long-run Neutrality',
National Bureau for Economic Research Discussion Paper, 6608.
Bernanke, B. and M. Woodford, (1996), Ìn¯ation Forecasts and Monetary Policy',
Princeton University, October, mimeo.
Blanchard, O., (1989), À Traditional Interpretation of Macroeconomic Fluctuations',
Clarida, R. and M. Gertler (1997) `How the Bundesbank Conducts Monetary Policy', in
C.Romer and D. Romer (eds), Reducing In¯ation, Chicago, Chicago University Press.
Clarida, R., J. Gali and M. Gertler (1997) `Monetary Policy Rules and Macroeconomic
Stability: Theories and Some Evidence', New York University, May, mimeo.
ÐÐÐÐ (1998) `Monetary Policy Rules in Practice', European Economic Review, 42, 1033±
68.
Cohen, D. and C. Wyplosz, (1989) `The European Monetary Union: An Agnostic
Evaluation', Centre for Economic Policy Research Discussion Paper, 306.
DeGrauwe, P. (1989) Ìs the European Monetary System a DM-Zone?', Centre for European
Policy Studies Working Document, 39.
Duck, N. W. (1988) `Money, Output and Prices: An Empirical Study Using Long-term
Cross Country Data' European Economic Review, 32, 1603±19.
ÐÐÐÐ (1993) `Some International Evidence on the Quantity Theory of Money',
Journal of Money, Credit and Banking, 25, 1±12.
Fisher, M. E. and Seater, J. J. (1993) `Long-Run Neutrality and Superneutrality in an
ARIMA Framework', American Economic Review, 83, 402±15.
Fuller, W. A. (1976) Introduction to Statistical Time Series, New York: John Wiley.
Fratianni, M. and J. von Hagen (1990a) `German Dominance in the EMS: Evidence from
Interest Rates', Journal of International Money and Finance, Vol. 9, pp. 358±375
ÐÐÐÐ (1990b) `The European Monetary System: Ten Years After', in A.H.Meltzer and
C. Plosser (eds), Carnegie±Rochester Conference Series on Public Policy, 32.
Gali, J. (1992) `How Well Does the IS-LM Model Fit the Postwar US Data?', Quarterly
Journal of Economics, may, 709Ð38.
Gerlach, S. and F. Smets (1998) Òutput Gaps and Monetary Policy in the EMU Area',
Bank for International Settlements, September, mimeo.
Geweke, J. (1982) `Measurement of Linear Dependence and Feedback Between Multiple
Time Series', Journal of the American Statistical Association, 77, 304±13.
ÐÐÐÐ (1986) `The Superneutrality of Money in the United States: An Interpretation of
the Evidence', Econometrica 54, 1±21.
King, R. G. and M. W. Watson (1997) `Testing Long Run Neutrality' Federal Reserve Bank
of Richmond, Economic Quarterly, 83, 69±101.
Kuttner, K. N. and A. S. Posen (1999) `Does Talk Matter After All? In¯ation Targeting and
Central Bank Behaviour', mimeo. CFS Working paper 99/04.
Lothian, J. R. (1985) Èquilibrium Relationships Between Money and Other Economic
Variables' American Economic Review, 75, 828±35.
Lucas, R. E. Jr, (1972) Èconometric Testing of the Natural Rate Hypothesis', In O.
Eckstein (ed.), `The Economics of Price Determination', Washington, DC, Board of
Governors of the Federal Reserve System.
ÐÐÐÐ (1976) Èconometric Policy Evaluation: A Critique. The Phillips Curve and
Labour Markets', Carnegie±Rochester Conference Series on Public Policy, 1, K. Brunner and
A. H. Meltzer (eds), Amsterdam: North-Holland.
158 Axel A. Weber
ÐÐÐÐ (1980) `Two Illustrations of the Quantity Theory of Money', American Economic
Review, 70, 1005±14.
McCallum, B. T. (1984) Òn Low-frequency Estimates of Long-run Relationships in
Macroeconomics', Journal of Monetary Economics, 14, 3±14.
Mills, T. C. (1982) `Signal Extraction and Two Illustrations of the Quantity Theory',
Mishkin, F. S. (1984) `The Real Interest Rate: A Multi-country Empirical Study', Canadian
Journal of Economics, 1, 283±311.
Orphanides, A. (1999) `The Quest for Prosperity Without In¯ation', Board of Governors
of the Federal Reserve System, May, mimeo.
Rudebusch, G. and L. Svensson (1998) `Policy Rules for In¯ation Targeting', Centre for
Economic Policy Research Discussion Paper, 1999, October.
Sargent, T. J. (1973) Ìnterest Rates and Prices in the Long Run: A Study of the Gibson
Paradox', Journal of Money, Credit and Banking, 5, 383±449.
Sims, C. A. (1972) Money, Income and Causality', Papers and Proceedings of the American
Economic Association, 62, 540±52.
Stock, J. H. (1991) `Con®dence Intervals for the Largest Unit Root in US Macroeconomic
Time Series', Journal of Monetary Economics, 28, 435±59.
Summers, L. H. (1983) `The Non-adjustment of Nominal Interest Rates: A Study of the
Fisher Effect in J.Tobin (ed.) Macroeconomics, Prices and Quantities: Essays in Memory of
Arthur Okun, Washington DC, Brookings Institution.
Taylor, J. B. (1993) `Discretion versus Rules in Monetary Policy Making', in A.H.Meltzer,
and Plosser, (eds), Carnegie Rochester Conference Series on Public Policy, 39, December,
195±214.
Weber, A. A. (1990) Èuropean Economic and Monetary Union and Asymmetries and
Adjustment Problems in the European Monetary System: Some Empirical Evidence',
Centre for Economic Policy Research Discussion Paper 448.
ÐÐÐÐ (1991) `Credibility, Reputation and the European Monetary System', Economic
Policy, 12, 57±102.
ÐÐÐÐ (1994) `Testing Long-run Neutrality: Empirical Evidence from the G7-
Countries with Special Emphasis on Germany', in A. H. Meltzer and Plosser (eds),
Wyplosz, C. (1991) ÈMS Puzzles', Revista Espanola de Economia, 7(1), 33±66.
Discussion
Carlo A. Favero
This Comment offers some criticisms of the empirical literature on monetary

policy rules by focusing on two main issues: interpretation of estimated
parameters and the effects of misspeci®cation.
On the issue of identi®cation it is argued that it is impossible to disentangle
the causes and consequences of policy actions in any policy reaction
function. Reduced form econometric methods are unable to discriminate
empirically between an interest rate policy reaction function (an anti-
in¯ation feedback rule) on the one hand and the in¯ationary consequences
of the interest rate policy on the other. The potential impact of simultaneity
is addressed by estimating bivariate VARs in ®rst differences for in¯ation and
interest rates to apply a methodology proposed by King and Watson(1997).
Such a methodology allows us to evaluate the results of all possible
identi®cation schemes in bivariate VAR models. Imposing that monetary
policy has long and variable outside lags ± i.e. in¯ation initially responds
moderately to interest rate changes ± and that monetary policy is neutral in
the long-run ± i.e. a long-run Fisher effect exists ± it is found that monetary
policy of the Bundesbank is unlikely to be adequately described by an interest
rate smoothing rule. In fact, a long-run unit effect of in¯ation on interest
rates for Germany is compatible with no short-run policy reaction of interest
rates to in¯ation at all. The same prior beliefs leads to the conclusion that for
the United States only a very large immediate response of interest rates to
in¯ation would be consistent with a long-run neutral monetary policy
stance. The author concludes that these results point clearly towards a
problem of lack of exogeneity in single-equation speci®cations for monetary
policy rules.
On the issue of misspeci®cation, it is argued that it is impossible to discuss
European interest rate policy rules in isolation.
I shall introduce the discussion by proposing ®rst a general framework to
understand identi®cation and speci®cation of monetary policy rules.
Within such a general framework, I shall address both general and speci®c
issues.
159
160 Carlo A. Favero
1 A general framework to understand monetary policy rules: the

simplest model of the monetary transmission mechanism
Identi®cation and speci®cation problems for monetary policy rules are best
understood by brie¯y outlining their derivation. I shall consider here one of
the simplest possible models, taken by Svensson (1998). The central bank faces
the following intertemporal problem:
X
1
Et i Lti
i0
1
L t � 2 D 4:1
2
where Et denotes expectations conditional upon the information set available
at time t , is the relevant discount factor, t is in¯ation at time t and is the
target level of in¯ation. The central bank is de®ned as a strict in¯ation targeter.
As the monetary instrument is the policy rate, it , the structure of the economy
must be described to obtain an explicit form for the policy rule. We consider
the following speci®cation for aggregate supply and demand in a closed
economy:
xt1 x xt � r it � Et t1 � r z zt udt1 D 4:2
t1 t x xt z zt ust 1 D 4:3
where x represents deviations of output from its natural level and is the rate
of in¯ation; zt is a vector of exogenous variables.
The following interest rate rule can be derived from the ®rst-order
conditions for optimality:

1 x r
it r Et t1 � D 4:4
x r
x z z
xt zt Et zt1
r r x r
A number of comments on D4:4 are in order:
. Fitted parameters are convolutions of the parameters describing central

banks' preferences( ) and those describing the structure of the economy
x ; r ; x ; r :
. The economy is best described by a three-equation model (under the
assumption of valid exogeneity of the variables included in zt : Identi®ca-
tion of parameters in the monetary policy rule is achieved by the block-
recursivity assumption: interest rates react simultaneously to all the variables
Discussion 161
in the economy, but the reaction of in¯ation and output to monetary

policy takes at least a one-period lag
. In order to make D4:4 consistent with the data, sluggish adjustment of
actual to desired policy rate has to be imposed (see Clarida, Gali and Gertler,
1998).
. Only one empirical implication can be confronted with the data by
estimating the monetary policy rule as a single equation ± namely, whether
the parameter describing the reaction of policy rates to a gap between
current and target in¯ation is larger than one.
2 Some considerations on identi®cation and speci®cation

problems of monetary policy rules
The two issues raised by Weber in Chapter 4 are easily understood within the
framework proposed in Section 1. The identi®cation problem can be referred
to the estimation of the monetary policy rule as a single equation, when the
data are generated by a multi-equation model. A related issue is the validity of
the block-recursivity assumption. In general, it is absolutely true that the
estimation of the full model is much more informative than a single-equation
speci®cation of the monetary policy rule. In fact, tests of the compatibility of
actual behaviour of central banks with different type of in¯ation targeting
models can be conducted only by checking cross-equation restrictions on
parameters of multi-equation models.
However, the two-equation speci®cation for in¯ation and interest rates
adopted in Chapter 4 is also a misrepresentation of the full model, in that it
leaves at least one equation out of the picture. Therefore, the validity of the
Weber critiques on identi®cation might be very limited if the omission of the
equation for output leads to invalid estimation and inference in the two-
equation model.
The point of misspeci®cation as a consequence of interdependence of
European interest rates is easily interpreted as a misspeci®cation of the vector
zt , in the case where the German interest rates are exogenous. If this does not
happen, then one more dimension has to be added to the problem by
considering in¯ation targeting in open economies. As a consequence, the full
model of the economy includes at least four equations, and misspeci®cation
problems induced by considering just two of them are likely to become more
severe.
Some speci®c issues on the econometric approach

Weber estimates bivariate VARs in ®rst differences for in¯ation and interest
rates using the methodology in King and Watson(1997) to assess the impact of
different identifying assumptions on tests of the relevant hypotheses.
We have already pointed out a potential misspeci®cation induced by
considering bivariate models. As well explained in Chapter 4, this choice is
162 Carlo A. Favero
driven by the `curse of dimensionality problem', which affects the methodo-

logy. Leaving aside this problem, it is probably worth considering a technical
issue related to the speci®cation of VARs in ®rst differences.
In general, the application of the King and Watson (1997) methodology
requires the ability to compute ®nite cumulative impulse responses on which
the long-run restrictions are imposed, and therefore the speci®cation of
stationary VARs. Stationarity is achieved in Chapter 4 by ®rst-differencing.
However, if series are cointegrated, then VAR in ®rst differences delivers biased
estimates of the parameters of interest. Interestingly, the long-run restrictions
considered in the two main sections of the Chapter are the Fisher effect and
the uncovered interest parity (UIP). Both relationships deliver cointegrating
vectors, between nominal interest rate and in¯ation and between domestic
and foreign interest rates, respectively. It might then be more appropriate to
achieve stationarity by considering an error-correction speci®cation.
�
Consider, for simplicity, the case of a bivariate model yt yt ; xt ; in which
variables are nonstationary I(1) but cointegrated with cointegrating
vector1; �1:

yt 11 yt�1 b11 b12 v1t
1 � 1 D 4:5
xt 21 xt�1 b21 b22 v2t
A VAR in ®rst differences is clearly misspeci®ed because it omits the variable

capturing the long-run relationship between x and y. However, model D4:5
can be rewritten as follows:

�1 1 1� L 0 yt � xt 11 0 yt�1 � xt�1

0 1 0 1 xt 21 0 xt�1

b b12 v1t
11 D 4:6
b21 b22 v2t
The two representation are absolute identical (same residuals). The cointegrat-
ing properties of the system suggests the presence of two types of shocks: a
permanent one (to be related to the single common trend shared by the two
variables) and a transitory one (to be related to the cointegrating relation). It
seems therefore natural to identify one shock as permanent, the other as
transitory. Given that we have a stationary system, the identi®cation of shocks
is obtained by deriving long-run responses of the variables of interest to
relevant shocks. From D4:6 we have:

�1 1 1� L 0 11 L 0 yt � xt
�
0 1 0 1 21 L 0 xt

b11 b12 v1t
D 4:7
b21 b22 v2t
Discussion 163
From which long-run responses are obtained by setting L 1 and by inverting

the matrix, premultiplying variables in the stationary representation of VAR:
�1
yt � xt �11 1 b11 b12 v1t
D 4:8
xt �21 1 b21 b22 v2t
!
�b11 b21
yt � xt 11 �21 -b12 �b22
11 �21
v1t
�21 b11 11 b21 -21 b12 11 b22 D 4:9
xt 11 �21 11 �21
v2 t
so v2t can be identi®ed as the transitory shock by imposing the following

restriction:
�21 b12 11 b22 0
which, given knowledge of the -parameters from cointegration analysis,

provides the just-identifying restriction for the parameters in B1 . This little
example illustrates that speci®cation in ®rst differences might be a source of
mis-speci®cation and that the presence of cointegration between the series of
interest leads naturally to a set of speci®c identifying restrictions, which could
be exploited to limit the effects of the curse of dimensionality.
To conclude, we agree with the main message of this chapter, which invites
us to consider monetary policy rules in the context of macromodels rather
than in isolation. Bivariate models speci®ed in ®rst differences are a ®rst step in
this direction; it would be desirable to extend such framework to achieve a
closer relation between theoretical and applied models and to avoid
misspeci®cation of applied models owing to the omission of long-run
equilibria.
Note
1. Interestingly there is one case in which such structural identi®cation is achieved
thorugh the standard Choleski decomposition: this is the case in which yt is
weakly exogenous for the estimation of b21 :
References
Clarida R., J. Gali and M. Gertler (1998) `Monetary Policy Rules in Practice:some
Svensson L. E. O. (1998) `Monetary Issues for the ESCB', paper presented at the Carnegie±
Rochester Conference; available at < http://www.iies.su.se/leosven/ >.
King R. G. and M. W. Watson (1997) `Testing Long-run Neutrality', Federal Reserve Bank
of Richmond Economic Quarterly, 83, 69±101.
Discussion
Philippe Moutot
Over the past few years, the estimation of policy reaction functions has
become quite popular. Nevertheless, studying empirical literature often seems
much like a `®shing expedition'.2 The Weber's study in Chapter 4 adds some
interesting results to an important issue, namely the issue of robustness.
The chapter argues that, on the one hand, it is not possible to disentangle
the determinants and consequences of monetary policy without prior
restrictions, a fact which is well known in econometrics as the ìdenti®cation
problem'. On the other hand, it is stated that it is impossible to analyse the
interest rate policy rules of the European countries in isolation, owing to the
interest rate linkages via the ERM. Both problems are not new and ± as far as I
can see ± unsolved in literature so far.
My comments will be threefold. I will ®rst argue that the kind of issues
tackled in this chapter and the techniques used for this purpose are
increasingly relevant, at least from the point of view of a central banker. I
will then discuss the results of the chapter to see what lessons might be learned
from it with regard to the identi®cation of rules, submitting that this chapter
should only be the ®rst step of a long-term research project. I would, however,
caution against the ultimate hope that rules could fully replace discretion in
the context of central banking.
1 Both the issues raised and the techniques used are relevant
The issues
In today's empirical literature, VAR and Vector Error Correction (VEC) models
have become increasingly popular tools. Unfortunately, and as is also well
known, such reduced form models per se do not provide us with any
knowledge about causality.
Following this line of reasoning, traditional estimates of the monetary policy
reaction ± e.g. those referred to as `Taylor Rules' ± cannot in most cases discriminate
between the reaction function ± e.g. monetary policy changes in response to in¯ation,
and the transmission process running from the interest rate to in¯ation, owing
164
Discussion 165
to the joint interaction of all variables being under investigation. Therefore,

the very use of such reaction functions by central bankers either as actual rules
± or, more realistically, as possible references for action ± is questionable. This
is all the more true given that, ex ante, the general public may have held
different views of the action of monetary authorities, implying that, once
implemented, the rule could actually lose its signi®cance, another case of
`Goodhart's Law'.
The issue is relevant not only in the case of VARs and VECs, but also in the
case of large econometric models, where the estimation of structural
parameters is often very much dependent on the underlying assumptions
concerning the reaction functions of authorities. Therefore, determining a rule
by optimising a loss function on the basis of a structural model, estimated without
adequately facing the òbservational equivalence' issue, is also problematic.
Finally, the European case is even more complex than the cases of other countries
in such respects. Owing to monetary union, the monetary regime changed
drastically at the start of 1999, so that the past reaction functions of monetary
authorities are an uncertain guide. Moreover, owing to the asymmetric
functioning of the EMS, an average of these reaction functions is likely to be of
uncertain interest, given that even the nature of variables within these
putative reaction functions may have varied across countries. Taking into
account the past nature of reaction functions in the EMS countries is therefore
advisable for any econometric analysis of Euro Area data which attempts to
draw inference of relevance to the European Central Bank (ECB). This is
certainly an issue which the Eurosystem itself considers today in the
construction of its econometric models.
In this context, a thorough investigation of the leadership role of the
Bundesbank in the determination of European interest rates before monetary union
is certainly relevant, given that this constitutes a necessary ®rst step to model
these monetary regimes, although the conclusions of the analysis have been
clear for a long time, at least to European central bankers.
In the next step of his research, I would encourage A. Weber to extend his
analysis towards an investigation of the relationships among in¯ation, the
short-term interest rate and other key economic variables such as output or the
slope of the yield curve in a simultaneous framework.
The techniques
One way of dealing with the issue is to apply the Choleski decomposition
which is widely used in the older literature and which requires an ordering of
the variables. Owing to its rather mechanical nature, this procedure is
increasingly being viewed as unsatisfactory.
Another possibility for overcoming this problem is to be found in a
structural approach ± i.e. in imposing restrictions either on the short-run or on
the long-run coef®cients in the variance±covariance matrix.3 Inference can
then be carried out in two ways: one way could be to choose an identi®cation
166 Philippe Moutot
scheme (in fact re¯ecting the researcher's a priori economic assumption about
the nature of the shocks and therefore also about causality) and subsequently
to derive the corresponding point estimates. Another way could be to impose
several identi®cation schemes and to deduce whether the values of the
parameters of interests are in line with them. The second alternative has been
used in the well-known study by King and Watson (1997),4 that provides a
rather ¯exible and elegant scheme for imposing and testing identifying
restrictions and which incidentally also coincides with the spirit of Edward
Leamer's early work.5
This is also the alternative chosen by Weber. In the end, Weber estimates its
parameters by imposing a restriction that is compatible both with economic
theory and with the data. Given the generally low power of the stationarity
tests used by Weber to test for co-integration in small samples, the use of VEC
rather than VAR models might have been considered.
2 Interpreting the results from a central banker's point of view
Following this line of reasoning, Weber investigates a number of interesting

hypotheses related to interest rate functions.
The long-run neutrality of money

In the ®rst part of Chapter 4 Weber addresses the questions of the long-run
neutrality of money (e.g. the hypothesis that nominal shocks have no
in¯uence on real variables) and the long-run homogeneity of prices (e.g. the
hypothesis that nominal interest rates and in¯ation must change proportio-
nately in response to a nominal shock, the `Fisher effect') for a number of
countries.
As a consequence of his setup, it turns out that, once in¯ation is
endogenised in the model and the usual identi®cation procedure is carried
out, the model is no longer identi®ed. In other words: an additional restriction
is needed to estimate the model. Weber therefore imposes a long-run
correlation of one between nominal rates and in¯ation and shows in the
next step that the Fisher effect is supported by data only under the assumption that
the short-term impact of in¯ation on the interest rate and vice versa belong to a
bounded set (see Figure 4.3). However, since the resulting parameter values are in
line with the ones given by the simple (unrestricted) VAR, the hypothesis
seems to be in line with the data.6 All in all, Weber does not reject the Fisher
effect for the United States, Germany, Canada and Japan.
Let me turn back, however, to the more general lessons that can be drawn
from this exercise.
First, indeed, the Fisher effect seems to be borne out by the data. However,
the estimation of the rule and of the corresponding structural model cannot be
made with great precision. In most cases, the hypothesis of a rule combined
with a well functioning economic structure is reasonable. But it cannot be
Discussion 167
identi®ed with full precision. For each assumption on the short-term impact of
in¯ation on interest rates, there is a corresponding separate estimate of the
interest rate rule.
Second, the hypothesis of a rule combined with a well functioning
economic structure cannot always be con®rmed. This is obviously the case with
the United Kingdom where some more detailed analysis needs to be carried
out. One may wonder if the time series is not too long in that case, mixing too
many different monetary regimes and therefore being misleading. Somehow,
con®rming a rule also depends on the actual period on which it can be
observed.
Third ± and this is meant to be a more philosophical conclusion ± it is not at
all sure whether it is possible for a new central bank like the ECB to ®nd optimal
rules èx ante'. Owing to this issue of òbservational equivalence', some
`learning by doing' is unavoidable. Credibility is of the essence in such a
process. And the way credibility in¯uences the functioning of the economy is
likely to depend on the nature of the rule itself. This is a point which becomes
even more clearly visible when considering the second type of issue tackled by
Weber.
The leadership of German interest rates in the EMS

Weber then tests for the long-run interactions between German interest rates
and the interest rates of other European as well as some G-7 countries by
taking the difference between the immediate and long-run transmission
explicitly into account once again. In sum, he ®nds evidence in favour of an
asymmetrical EMS with Germany as the anchor country. This can be seen
from the long-run interest rate effects (see Table 4.5, p. 152).
Concerning the case of France, however, I would like to add some
considerations that might complement the interpretation slightly. Owing to
the EMS constraint, French short-term interest rate changes have indeed
generally followed German ones in the short run. However, as the credibility of
the exchange rate peg improved, it was in some occasions possible for the
Banque de France to avoid or only partially follow some upward movements of
German rates or to exceed some of their downwards movements. Thus the
French short-term rate would converge gradually to the German one. However,
this was possible only when the economic situation ± and, in particular, the
growth of output ± made it clear that the French authorities would not be faced
with any serious dilemma in the case of a speculative attack on the EMS.
Moreover, when growth was expected to dip, speculative attacks could ¯are up
leading the short-term rates and, at the start of the 1980s, also the long-term rates
to higher levels. Therefore, it might be expected that the slope of the French yield
curve would show a more direct link to the output gap than the German one. As
mentioned above, extending the system to additional variables like output and
the yield curve could provide us with additional insights.
Again, the credibility of monetary policy seems to have been of the essence.
168 Philippe Moutot
3 Rules can limit, but not replace discretion
To conclude, progress can be made in a number of areas to better identify rules

useful to monetary authorities. In Chapter 4 Weber has initiated a useful research
programme and is trying to tackle essential issues. But the hope that rules can, as
a result of such analysis, replace discretion should not be overestimated.
Going beyond the problems of òbservational equivalence' mentioned
above, it should be recognised that the rules estimated on the basis of central
banks' past behaviour usually contain an auto-regressive pattern. In other
words: without adding, though usually without a strong economic justi®ca-
tion, an auto-regressive part to the in¯ation and output gap variables, a `bad
®t' would result. Simultaneously, rules deduced from the optimisation of a loss
function applied to a structural model usually yield optimal reaction
functions, which create an excessive degree of interest rate volatility.
Therefore, a good explanation to the inertial behaviour of central banks is
needed. A possible explanation is given by Woodford (1998), who indicates
that an optimal central bank should be able to commit itself over the long term
to best in¯uence long-term rates, thereby needing inertia. This is in line with
earlier remarks by Goodfriend (1991). Moreover, it is consistent with the
conclusion I drew earlier, according to which credibility and its evolution are
essential variables which must be taken into account.
However, credibility is always the combination of two elements. Credibility
entails, ®rst, doing what one promised to do. But it also implies that what is
done is generally viewed as reasonable by the public.
Ful®lling these two conditions in a changing world, the economic model of
which is both changing and only partially known, implies the following: while
making good on its commitments, the central bank needs to adapt its view of the
economy and its explanation of the future, whenever needed. While the analysis
of commitments is increasingly well studied by academic literature, the analysis
of this discretionary need to adapt models and views is usually ignored. However,
while continuously listening to new ideas developed by academics and market
people, the central bank needs to pass judgement on whether such new views
should be integrated in its analysis. This is essential to offer the best monetary
policy possible to the public. The central bank therefore needs discretion to
adapt its analysis to new developments. In view of current uncertainty on the
way both the global economy and the Euro Area economy function, such a need
for discretion should not be overlooked.
A ®xed monetary policy rule is therefore not reasonable. However, the
relevant issue in this context is how to combine the need for some limited
discretion with the need for transparency vis-a-vis the public.
The best solution for a central bank facing this issue is, in my view, to use one
or several rules as reference points for its action and to endeavour to explain to
the public the deviations from such rules in a systematic manner. In this view,
describing the particular changes in indicators or in views that motivate such
Discussion 169
deviations seems to be at least as important for transparency as providing the

overall view of the economic prospects that synthesise the decisions made.
This is what the ECB has chosen to do. By mentioning that the ®rst pillar of
its strategy is a reference value for a monetary aggregate, it made clear its strong
prior conviction that in¯ation is always, in the end, a monetary phenomenon.
Simultaneously, by indicating that its strategy has a second pillar ± the outlook
for price developments ± it also showed that the model of the economy it has
in mind may evolve over time and that it would also use a wide set of analyses
and indicators, including forecasts, to motivate its decisions. Further, its desire
to ®nd the basis for this outlook on both indicators and forecasts without
giving an excessive role to such forecasts is consistent with the need to give the
best possible justi®cation to and the best possible account of the unavoidable
discretion entailed by good monetary policy making.
Notes
1. Prepared by Philippe Moutot, with the help of Dieter Gerdesmeier.
2. Shiller (1990), p. 667. Shiller was actually referring to the estimation of the
determinants of the term structure, but his statement seems to be valid for our
considerations as well.
3. See Sims (1986) and Bernanke (1986) for example.
4. King and Watson have also used the framework for the estimation of the long-run
Phillips Curve (i.e. the relation between in¯ation and unemployment) which is of
particular interest for a central bank.
5. See Leamer (1978) and (1983) for a more detailed overview.
6. It should be stressed that a simple correlation between the residuals in the simple
VAR can easily be calculated. However this does not say anything about the
causality. This question cannot be answered without imposing identifying
restrictions and testing for them.
References
Bernanke, B. (1986) Àlternative Explanations of the Money-Income Correlation',
Goodfriend, M. S. (1991) Ìnterest Rates and the Conduct of Monetary Policy', (Carnegie±
Rochester Conference Series on Public Policy, 34, 7±30.
King, R. G. and M. W. Watson (1997) `Testing Long-run Neutrality,' Federal Reserve Bank
of Richmond Quarterly, 83(3), 69±101.
Leamer, E. E. (1978) Speci®cation Searches: Ad hoc Inference with Non-experimental Data,
New York: John Wiley.
ÐÐÐÐ`Let's Take the Con out of Econometrics', American Economic Review, 73, 31±43.
Shiller, R. J. (1990) The Term Structure of Interest Rates, in [Friedman, B. M., and Hahn,
F. H. (eds.), Handbook of Monetary Economics, vol. I, Elsevier Science Publishers
B.V., pp. 627±722., 1990].
Sims, C. (1986) Àre Forecasting Models usable for Policy Analysis?', Federal Reserve Bank
of Minneapolis Quarterly Review, 10, 2±16.
Woodford, M. (1998) Òptimal Monetary Policy Inertia', paper presented at the CFS
conference in Frankfurt, October.
5
Legal Structure, Financial Structure

and the Monetary Policy Transmission
Mechanism1
Stephen G. Cecchetti
1 Introduction
Over the past decade, the countries of central Europe have become more alike
in many ways. As the new members of the European Monetary Union (EMU)
prepared for the birth of the Euro on 1 January 1999, their economic policies
became substantially more uniform. All eleven countries in the new Euro Area
had virtually eliminated in¯ation and taken serious steps toward ®scal
consolidation.2 As their monetary and ®scal policies have adjusted to meet
these common goals, the countries' business cycle ¯uctuations appear to have
become more synchronised as well.3 While this makes the job of the European
System of Central Banks (ESCB) easier, numerous dif®cult challenges remain.
Primary among these is the making of policy in the face of the possibility that
it will have differential impacts across the countries of the Euro Area.
The task facing the ESCB is even more complex than that facing countries with
stable monetary regimes, where the measurement of the national and regional
impact of policy has already proved to be extremely dif®cult. The creation of the
ESCB constitutes a `regime shift' in virtually every sense of the term. The
introduction of the Euro seems sure to prompt adjustments in the economies of
the member countries, and these adjustments will probably alter the relationship
between the actions of the central bank and the real economy. That is, the
monetary transmission mechanism of the countries in the Euro Area will change,
making the job of the new European Central Bank (ECB) even more dif®cult than
it is already. But how quickly will it change, and what will it become?
To answer these questions, we must understand the fundamental determi-
nants of the impact of policy actions on output and in¯ation. For insight into
these determinants, I turn to the modern views of the monetary transmission
mechanism, which assign a central role to ®nancial structure. Kashyap and
Stein (1997) provide a starting point; they focus on the importance of the
banking system and go on to emphasise the distributional effects of monetary
policy changes. The conventional wisdom has always been that some
industries are more sensitive to interest rate changes than others, and so
170
Legal Structure, Financial Structure and Monetary Policy Transmission 171
changes in policy-controlled interest rates have differential effects across

industries. The view based on ®nancial structure both formalises this
reasoning and takes it one step further by noting that some ®rms are more
dependent on banks for ®nancing than others, and that this is true both across
and within industries. According to this `lending view' of the transmission
mechanism, monetary policy actions change the reserves available to the
banking system, thereby affecting the willingness of banks to lend ± and,
ultimately, the supply of loans. How this mechanism will affect individual
®rms depends on the ®nancing methods available to them. Monetary policy
has a bigger impact on ®rms that are reliant on banks for their ®nancing.
Furthermore, healthier banks will be able to adjust to the policy-induced
reserve changes more easily than other banks will.
The distributional effects implied by the lending view of monetary policy
transmission have clear implications for the Euro Area and the ESCB.
Countries in which ®rms are more bank dependent and banking systems are
less healthy will be more sensitive to the Eurosystem's decisions to change
interest rates. This brings me to the ®rst question I will address in this chapter:
is there evidence that the impact of monetary policy innovations varies across
countries with the strength and scope of the banking system?
With this in mind, I examine differences in the size, concentration and
health of national banking systems, as well as in the availability of non-bank
sources of ®nance. I ®nd, consistent with the most casual observation, that
banking system characteristics vary dramatically across the countries of the
European Union (EU). Furthermore, these differences do seem to be related to
estimated differences in the impact of interest rate changes on output and
in¯ation. Countries with many small banks, less healthy banking systems and
poorer direct capital access display a greater sensitivity to policy changes than
do countries with big, healthy banks and deep, well developed capital markets.
But this is just the ®rst question. The more important issue facing the ESCB
is whether the national banking systems, and the implied sensitivity of each
country's real economy to monetary policy shocks, will change now that there
is monetary union.
It is easy to assert that European banks will soon look like US banks,
exhibiting a ®nancial structure and transmission mechanism similar to
American models. After all, the Euro Area does resemble the United States,
at least super®cially. It has a slightly larger population ± 292 million for the
eleven members of the monetary union relative to 270 million for the United
States ± and nearly as high a level of GDP ± $6.8 trillion compared with $8.1
trillion in 1997. The Euro Area also has a similar degree of openness to trade,
with imports accounting for slightly more than 10 per cent of GDP. These
parallels, along with the fact that ®nancial technology is easily transferable
across national boundaries, have led a number of observers to conclude that
the introduction of the Euro may act as a catalyst, speeding the rate at which
®nancial relationships in Europe become like those in the United States. For
172 Stephen G. Cecchetti
example, while Dornbusch, Favero and Giavazzi (1998, pp. 48±9) do note the
possibility for EU-wide asymmetries resulting from differences in ®nancial
structure, they assert that `the euro will change the way ®nancial markets
work, inducing corresponding changes in the monetary mechanism. In
addition to pervasive deregulation already under way and innovation, the
introduction of the euro will revolutionize the ®nancial structure of Europe.
Europe will in a short period become more nearly like the USA.' McCauley and
White (1997, p. 17) suggest that there may be an acceleration in the rate at
which securities replace loans on the asset side of bank balance sheets and
commercial paper replaces deposits on the liability side. They point to a
`dramatic potential for assets to be stripped out of the banking system' and for
securities markets to absorb as much as one-third of the corporate loans now
originated in European banks.4 Overall, these commentators are speculating
that the increased liquidity of European ®nancial markets brought about by
monetary union will lead to signi®cant consolidation of banks, with mergers
at both the national and the international level, as well as a direct substitution
of traded equities and bonds for bank loans.
Why should we believe that the European ®nancial structure will quickly be
transformed into one that mirrors the one in the US model? Without an
explanation for the evolution of these countries' national ®nancial structures
that is based on their existing differences, such claims are unconvincing. What
accounts for the variation in the ®nancial intermediation systems across
countries? Traditionally, we look to taxes and regulation for an explanation,
and Dornbusch, Favero and Giavazzi (1998) as well as White (1998) do
mention these. Danthine et al. (1999) identify a number of barriers to change
in national ®nancial structures and note the importance of the historical path
that has brought each country's banks to their current state. Danthine et al.
(p. 45) then go on to assert that `legal differences between EU states, in
particular the lack of some form of `Èuropean corporate law'', also remain
important and constitute an additional factor of market segmentation'. Such
disparities in legal structure can explain important economic patterns, and
they can be maintained for long periods of time, signi®cantly delaying the
harmonisation of national banking systems.5
It is my main contention that the differences in ®nancial structure across
the countries of Europe are a consequence of their dissimilar legal structures.
My argument draws on the work of La Porta et al. (1997, 1998), who focus on
the relationship between legal structures and ®nance. They argue that the
structure of ®nance in a country depends on the rights accorded shareholders
and creditors by the laws of that country, as well as on the degree to which
these laws are enforced. The nature of the laws is, in turn, a product of the legal
tradition on which the civil codes of a country are based. La Porta et al.
establish that the character of a country's ®nancial markets depends on the
country's legal structure. Putting their arguments together with the lending
view of the monetary transmission mechanism leads to the possibility that it is
Table 5.1 Effectiveness of 14cy and the origins of the legal system
Impact of policy
Index of effectiveness
Legal family of monetary policy On output on in¯ation
English 1.1 ±0.45 ±0.21

Scandinavian 1.8 ±0.52 ±0.22
French 2.1 ±0.70 ±0.20
German 2.4 ±1.25 ±0.49
Notes:
Index of effectiveness of monetary policy, from Table 5.5, is based on ®nancial structure variables
described in Section 3 of the text, with higher values implying a larger expected impact of interest
rate changes on output and prices.
The impact of policy on output and in¯ation, from Table 5.6, is a measure of the maximum
response, in percentage points, to an interest rate movement of 100 basis points, estimated using a
small-scale structural model.
Countries are classi®ed by the origin, or family, of their legal structure, and group means are reported
based on data for Ireland, the United Kingdom and the United States (English common law);
Denmark and Sweden (Scandinavian common law); Belgium, France, Italy, Portugal, and Spain
(French civil law); and Germany (German civil law).
the legal system in a country that forms the basis for the structure of ®nancial
intermediation and, hence, for the impact of monetary policy on output and
prices.
Table 5.1 reports the empirical ®ndings that support the basic conclusion of
the chapter. After classifying countries by the origin, or `family' of their legal
structure, I calculate for each family the average level of an index of monetary
policy's likely effectiveness (based on banking system size, concentration and
health, with a higher value implying greater effectiveness) and the estimated
impact of an interest rate change on output and in¯ation (from a small-scale
structural model). The results suggest that a country's legal structure, ®nancial
structure and monetary transmission mechanism are interconnected. The
clear pattern is that the predicted effectiveness and its measured impact vary
systematically based on the origin of a country's legal system. Countries with
better legal protection for shareholders and debtors (countries with a legal
structure based on English common law) have ®nancial structures in which
the lending channel of monetary transmission is expected to be less potent;
for these countries, the measured impact of an interest rate change on output
and in¯ation is lower.
The implication is that unless the laws governing shareholder and creditor
rights and the enforcement of those laws are harmonised across the members
of EMU, monetary policy will continue to have a differential impact. Put
slightly differently, it is my belief that the ®nancial structures in the countries
of the Euro Area will not converge into one large US-style system unless there
are dramatic legislative changes. If such legal harmonisation occurs ± that is, if
the civil codes protecting shareholders and creditors are made uniform across
the countries that have entered the monetary union ± then the regional
variation in the impact of interest rate changes on output and in¯ation should
decrease.6 But if legal convergence does not occur, ®nancial structure will
remain heterogeneous, and so will the monetary transmission mechanism,
and the job of the Eurosystem will be to construct an appropriate policy that
takes these asymmetries into account.7
The remainder of this chapter provides the building blocks for this
argument. In Section 2, I provide a brief survey of the theories of the
monetary transmission mechanism, focusing on the importance of ®nancial
structure to an understanding of monetary transmission. Section 3 assesses the
national banking systems, including measures of overall size, concentration,
health and the relative importance of non-bank ®nance. Overall, this analysis
allows me to evaluate the likely strength of the lending channel across
countries. Section 4 reports estimates, for a set of ten countries, of the impact
of an interest rate increase on output and in¯ation. These estimates follow the
pattern that is expected: countries where ®nancial structure data suggest that
the lending channel should be strong exhibit more sensitivity to monetary
policy movements. In Section 5, I present the data and arguments from La
Porta et al. (1997, 1998) on the relationship between legal and ®nancial
structures. This allows me to test the prediction that countries with poor
shareholder and creditor protections and poor law enforcement will have less
developed ®nancial systems and greater sensitivity of output and in¯ation to
interest rate changes. While far from being de®nitive, the results are consistent
with my main hypothesis: differences in legal systems give rise to variations in
national ®nancial structures, and these variations in turn lead to divergences
in monetary transmission mechanisms. So long as the legal systems of the
Euro Area countries remain distinct, the impact of interest rate changes across
these countries will differ.
2 Theories of the transmission mechanism
A number of comprehensive surveys of the theories of the monetary

transmission mechanism have appeared in recent years. These include Bernanke
(1993), Gertler and Gilchrist (1993), Kashyap and Stein (1994, 1997), Hubbard
(1995), and my own survey, Cecchetti (1995). As a result, I will be brief.
All theories of how interest rate changes affect the real economy share a
common starting point. A monetary policy action begins with a change in the
level of bank reserves. For this to have any real effects at all, there must be
nominal rigidities in the economy. Otherwise, a change in the nominal
quantity of outside money cannot have any impact on the real interest rate.
While the ability of the central bank to change the level of bank reserves is not
in question, the source of the nominal rigidity that allows the change in
reserves to alter short-run real rates of return has been under debate for
decades. The current state of this discussion is well summarised by Christiano,
Eichenbaum and Evans (1997). They distinguish three sets of theories: one set
based on sticky wages, a second based on sticky prices and a third built on the
idea of limited participation. The sticky-wage and sticky-price models, which
are the most familiar, rest on the idea that there are costs to nominal price and
wage changes, and so adjustments are infrequent. In limited participation
models, introduced in Rotemberg (1984), individuals (households) are unable
to adjust their cash balances suf®ciently rapidly in response to changes in the
environment ± that is, households have a limited ability to participate in
®nancial markets, and so must commit themselves to certain portfolio
holdings for relatively long periods of time.8
The sources of nominal rigidities are relatively unimportant for the
discussion of the mechanism by which interest rate changes have short-run
effects on output and prices, and so I will move directly to a discussion of the
current theories of the transmission mechanism.9 Our current views are based
on the work of Bernanke (1983), Bernanke and Blinder (1992) and Bernanke
and Gertler (1989, 1990). These authors distinguish between the traditional
money view, in which interest rate movements affect the level of investment
and exchange rates directly, and the lending view, in which ®nancial
intermediaries play a prominent role in transmitting monetary impulses to
output and prices. I will describe each of these in turn.
The traditional view, which is largely the foundation for the textbook IS ±
LM model, is based on the notion that reductions in the quantity of outside
money raise real rates of return. This outcome has two effects, the ®rst directly
from interest rates to investment and the second through exchange rates. An
interest rate increase reduces investment, as there are fewer pro®table projects
available at higher required rates of return. A policy action induces a
movement along a ®xed marginal-ef®ciency-of-investment schedule. This
interest rate channel will be more powerful the less substitutable outside
money is for other assets. The exchange rate channel is also familiar from
textbook models. Here, an interest rate increase results in a real appreciation of
the domestic currency, reducing the foreign demand for domestically
produced goods. Regardless of whether the transmission mechanism occurs
through the interest rate channel or the exchange rate channel, there is no real
need to discuss banks. In fact, there is no reason to distinguish any of the
òther' assets in investors' portfolios. This is a simple two-asset model.
An important implication of this traditional model of the transmission
mechanism concerns the incidence of the investment decline. Since there are
no externalities or market imperfections, only the least socially productive
projects ± those with the lowest rates of return ± go unfunded. As a result, the
capital stock is marginally lower, but, given that a decline is going to occur, the
allocation of the decline across sectors is socially ef®cient.
As most of the surveys cited earlier emphasise, the lending view has two
parts ± one that focuses on the impact of policy changes on borrower balance
sheets and another that focuses on bank loans. In both, the effectiveness of
policy depends on capital market imperfections that make it easier for some
®rms to obtain ®nancing than others. Information asymmetries and moral
hazard problems, together with bankruptcy laws, mean that the state of a
®rm's balance sheet has implications for its ability to obtain external
®nance.10 By reducing expected future sales and by increasing the cost of
rolling over a given level of nominal debt, policy-induced increases in interest
rates (which are both real and nominal) cause a deterioration in the ®rm's net
worth. Furthermore, there is an asymmetry of information in that borrowers
(®rms) have better information about the potential pro®tability of investment
projects than do creditors (banks). As a result, as the ®rm's net worth declines,
the ®rm becomes less creditworthy because it has an increased incentive to
misrepresent the riskiness of potential projects ± an outcome that will lead
potential lenders to increase the risk premium they require when making a
loan. The asymmetry of information makes internal ®nance of new
investment projects cheaper than external ®nance.
More important for the transmission mechanism per se is that some ®rms are
dependent on banks for ®nance and that monetary policy affects bank loan
supply. A reduction in the quantity of reserves forces a reduction in the level of
deposits, which must be matched by a fall in loans. Nevertheless, lower levels
of bank loans will have an impact on the real economy only insofar as there are
®rms without an alternative source of investment funds.
Substantial empirical evidence supports the importance of both capital
market imperfections and ®rm dependence on bank ®nancing. Kashyap and
Stein (1997) provide a summary of two types of studies. The ®rst type suggests
that banks rely to a large extent on reservable-deposit ®nancing and that, for
this reason, a contraction in reserves will prompt banks to contract their
balance sheets, reducing the supply of loans. The second type establishes that
there are a signi®cant number of bank-dependent ®rms that are unable to
mitigate the shortfall in bank lending with other sources of ®nance. Overall,
recent research does imply the existence of a lending channel.11
Models of monetary policy transmission based on ®nancial structure suggest
a natural place to begin looking for sources of cross-country differences in the
monetary transmission mechanism. The prediction is that, overall, the
transmission mechanism will be stronger in those countries where ®rms are
more bank dependent, and where the banking system is less healthy and less
concentrated. In the ®rst instance, ®rms that have less direct access to capital
markets are unable to blunt the effect of a contraction in bank loans. In the
second, banks themselves have restricted access to non-reservable deposits and
are forced to contract their balance sheets by more for a given change in
policy. In Section 3, I examine data on national ®nancial structure and try to
rank countries based on the likely strength of the transmission mechanism. To
the extent that these cross-country differences are present, then the lending
view implies that they will persist until the ®nancial structures become more
uniform.12
3 Likely strength of the transmission mechanism
In assessing the likely impact of an interest rate change on output and prices
in the various countries of the EMU, I follow the work of Kashyap and Stein
(1997) and assemble data on the size and concentration of the banking
systems, along with measures of banking system health, the importance of
bank ®nancing and the size of ®rms. The indicators are chosen to conform as
closely as possible to the economic quantities that the lending view suggests
should be important. The balance sheets of large, healthy banks are not as
sensitive to policy, because reserve contractions can be readily offset with
alternative forms of ®nance that do not attract reserve requirements. In
addition, I examine measures of the development of equity and debt markets
in the EMU countries. Firms with ready capital market access, which are more
likely to be found in countries with extensive secondary securities markets,
will be better insulated from bank loan-supply contractions. Combining these
measures, I construct an index of the probable strength of the monetary
transmission mechanism.13
Table 5.2 Size and concentration of the banking industry, by country, 1996
No. of Banks per Concentration ratios:

credit institutions million people top ®ve banks
Country (1) (2) (3)
Monetary union members

Austria 1019 126 48
Belgium 140 14 57
Finland 350 68 78
France 1373 24 40
Germany 3517 43 17
Ireland 62 18 41
Italy 937 16 25
Netherlands 172 11 79
Portugal 51 5 76
Spain 313 8 44
Members of the EU not in EMU

Denmark 117 22 17
Greece 20 2 71
Sweden 124 14 90
United Kingdom 478 8 28
Other countries
Japan 556 4 30
United States 10 803 40 17
Note: Cobcentration ratios are calculated as the percentage of each country's bank assets accounted
for by the ®ve largest banks.
Sources: See the Data Appendix (pp. 189±90).
To assess the importance of small banks in a country's ®nancial system,

Table 5.2 reports the number of banks, the number of banks per million
population, and measures of concentration for all of the EU countries plus
Japan and the United States.14 The data reveal that Austria and Finland have
many more banks per capita ± 126 and 68 per million people, respectively ±
than any of the other countries. The remaining countries fall into roughly
three groups: the United Kingdom, Japan and the southern European
countries of Spain, Portugal and Greece have less than 10 banks per million;
the United States and Germany have 40 or slightly more; and the remaining
countries have between 13 and 25.
Turning to the concentration measures in column (4) of the table, it is
interesting to note that countries with more banks do not necessarily have less
concentrated banking systems. France, for example, with 1373 banks and just
under 60 million people, has a fairly high concentration ratio: the top ®ve
French banks account for a sizable 40 per cent of total banking system assets
and the top ten for nearly two-thirds. Overall, Denmark and Germany have
the least concentrated banking systems in Europe. By contrast, large banks
clearly dominate Sweden, Finland, Belgium, the Netherlands and Greece. The
remaining countries are somewhere in between.
What do these ®ndings imply for the strength of the transmission
mechanisms in the countries examined? Austria, Germany and the United
States have systems composed of a network of small banks, and so one would
expect the lending channel to be relatively strong in those countries. At the
other end of the spectrum, Belgium, Finland, Ireland, the Netherlands,
Portugal, Sweden and the United Kingdom all have banking industries
dominated by a small number of relatively large banks, with a modest
periphery of small institutions. The remaining countries ± Denmark, France,
Greece, Italy and Japan ± fall in a middle group.
The weaker a nation's banking system, the stronger the expected impact of
policy movements. With this in mind, I have collected a set of standard gauges
of banking system health ± return on assets, loan loss provisions, net interest
margin and operating costs ± and I have calculated a summary rating of overall
system soundness (Table 5.3). Focusing primarily on the return on assets and
the average Thomson ratings in Table 5.3 leads to the following rankings:
Ireland, the United Kingdom and the United States have the healthiest banks;
Austria, Belgium, Germany, the Netherlands, Spain, Denmark and Greece are
second; Finland, France, Italy, Portugal and Sweden are third; and Japan is
alone at the bottom.
Finally, I turn to the availability of non-bank ®nance for ®rms in EU and
other countries. The relevant data are reported in Table 5.4. Following Kashyap
and Stein (1997) and La Porta et al. (1997), I examine the number of publicly
listed ®rms, the extent of secondary equity and debt markets and the ratio of
bank loans to all forms of ®nance. Although these are crude measures of access
to external ®nance, they are informative. As in the case of Table 5.2, the
Table 5.3 Measures of banking industry health, by (country, 1996, per cent
Return Loan loss Net interest Operating Average Thomson

on assets provisions margin costs rating
Country (1) (2) (3) (4) (5)

Austria 0.38 0.59 1.67 2.45 2.38 (4)
Belgium 0.52 0.17 1.41 1.67 2.00 (6)
Finland 0.50 0.78 2.07 3.05 2.83 (3)
France 0.36 0.24 1.43 1.84 2.28 (16)
Germany 0.44 0.18 1.24 2.19 1.97(19)
Ireland 1.57 0.17 3.36 3.32 1.83 (3)
Italy 0.33 0.62 2.32 3.19 2.57 (15)
Netherlands 0.75 0.26 2.06 2.48 2.10 (5)
Portugal 0.91 0.42 2.60 3.80 2.30 (5)
Spain 0.76 0.32 2.20 2.69 1.79(11)

Denmark 0.91 0.11 1.28 0.97 2.33 (3)
Greece 1.11 0.18 1.98 2.77 2.50(6)
Sweden 1.28 0.25 1.90 1.77 2.50 (5)
United Kingdom 1.28 0.18 2.15 2.42 2.04(23)
Other countries
Japan 0.01 0.75 1.17 1.03 3.32(44)
United States 1.42 0.10 2.68 3.51 1.73 (344)
Notes: Except for the Thomson ratings, all ®gures in the table are calculated as a percentage of total
bank assets. In column (5), the number of banks rated by Thomson in each country and used to
compute the average appears in parentheses.
Sources: See the Data Appendix (p. 190).
countries can be divided into three groups. Austria, Ireland, Italy, Portugal and
Greece appear to have the least well developed external capital markets. They
have small equity and bond markets, and bank loans account for a high
percentage of ®rm ®nancing. By contrast, Belgium, Denmark, Sweden, the
United Kingdom and the United States all have substantial secondary capital
markets, and banks are a less important source of ®nance. The remaining six
countries are somewhere in between these two groups.
Table 5.5 summarises the material in Tables 5.2±5.4 and suggests the overall
relative strength of monetary policy in the fourteen EU countries, Japan and
the United States. The ®nal column, `Predicted effectiveness of monetary
policy,' reports a measure of the effects of monetary policy on output and
in¯ation, where higher values suggest a stronger lending channel and
therefore a larger impact. Overall, the pattern is very similar to the one
reported in Kashyap and Stein (1997, Table 6). Most important, the predicted
effects of interest rate movements vary greatly across countries. For example,
looking at the EMU countries, one would expect that a given interest rate
Table 5.4 Importance of external and bank ®nance, by country, 1996
Publicly Market Corporate Bank loans as

Number traded capitalisation debt as a a percentage
of publicly ®rms per as a percentage percentage of all forms of
traded ®rms capita of GDP of GDP ®nance
Country (1) (2) (3) (4) (5)

Austria 106 13.15 15 46 65
Belgium 139 13.68 45 60 49
Finland 71 13.87 50 34 39
rance 686 11.75 38 49 49
Germany 681 8.32 29 58 55
Ireland 76 21.59 18 13 80
Italy 217 3.78 21 37 50
Netherlands 217 13.97 96 48 53
Portugal 158 16.11 23 19 62
Spain 357 9.09 42 11 58

Denmark 237 45.06 41 105 25
Greece 245 23.44 20 3 48
Sweden 229 25.90 99 73 32
United Kingdom 2 433 41.39 150 45 37
Other countries
Japan 2;34 18.56 67 39 59
United States 8,479 31.94 111 64 21
Notes: Market capitalisation is the year-end value of ®rms listed on major exchanges. For the United
States, three exchanges are used; for Japan, eight; and for each of the remaining countries, one.
Sources: See the Data Appendix (p. 190).
change would have the most impact in Austria and Italy, countries in which
small banks are relatively important, the banking systems are less healthy and
®rms have little access to non-bank sources of ®nance. The opposite is true of
Belgium, Ireland and the Netherlands, where the banking systems are large
and healthy and non-bank ®nance is readily available; in these countries,
interest rate movements would be expected to have a more muted impact.15
The conclusions of this section could be criticised as applying only to the
pre-EMU period. But will the introduction of the Euro be a catalyst for the
harmonisation of ®nancial structure across the EMU? I take this question up in
more detail later, but at this point I will simply mention that the European
Central Bank (1999) report, Possible Effects of EMU on the EU Banking Systems in
the Medium to Long Term, provides very little evidence to suggest that an
increase in either international banking competition or securitisation and
disintermediation will occur quickly.
Table 5.5 Summary of factors affecting the strength of the monetary transmissions
mechanism
Availability Predicted
Importance Bank of alternative effectiveness
of small banks health ®nance of monetary policy
Country (1) (2) (3) (4)

Austria 3 2 3 2.67
Belgium 1 2 1 1.33
Finland 1
3 2 2.00
France 2
3 2 2.33
Germany 3
2 2 2.33
Ireland 1
1 3 1.67
Italy 2
3 3 2.67
Netherlands 1
2 2 1.67
Portugal 1
3 3 2.33
Spain 2
2 2 2.00

Denmark 2 2 1 1.67
Greece 2 2 3 2.33
Sweden 1 3 1 1.67
United Kingdom 1 1 1 1.00
Other countries
Japan 2 4 2 2.67
United States 3 1 1 1.67
Notes: Column (1) is based on Table 5.2; column (2), on Table 5.3; and column (3), on Table 5.4.
Column (4) is an average of columns (1), (2) and (3).
4 Measuring the impact of policy on output and prices
Testing the proposition that the banking system's concentration, health and
importance have a material impact on the monetary transmission mechanism
requires an estimate of the effects of an interest rate change on output and
prices. Numerous studies report such estimates for some or all of the countries
of the EU. These include Gerlach and Smets (1995), who estimate a three-
variable structural vector autoregression based on long-run restrictions; de
Bondt (1997), who presents estimates of the impact of policy on output and
prices for Germany, France, Italy, the United Kingdom, Belgium and the
Netherlands that are based on the work of other authors; Dornbusch, Favero
and Giavazzi (1998), who report estimates of the impact of policy on output
and prices derived from both small vector-autoregressive models and large
structural models for Italy, Germany, France, Spain, Sweden and the United
Kingdom; Kieler and Saarenheimo (1998), who study France, Germany and
Per cent Per cent

1.0 1.0
Belgium France
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0 1.0
Germany Ireland
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0 1.0
Italy Portugal
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0 1.0
Spain Denmark
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0 1.0
Sweden United Kingdom
0.5 0.5
0 0
–0.5 –0.5
–1.0 –1.0
–1.5 –1.5
2 4 6 8 10 12 14 16 18 20 22 24 26 2 4 6 8 10 12 14 16 18 20 22 24 26
1.0
United States
0.5
0 Output response
–0.5 Inflation response
–1.0
–1.5
2 4 6 8 10 12 14 16 18 20 22 24 26
Figure 5.1 Reaction of output and in¯ation to an interest rate increase of 100 basis
points, quarterly, by country
Sources: Cecchetti (1996); Cecchetti, McConnell and Perez Quiros (1999).
the United Kingdom, concluding that the transmission mechanism is not

signi®cantly different across the three countries; Vlaar and Schuberth (1998),
who examine money demand functions for fourteen EU countries; Ehrmann
(1998), who estimates structural vector autoregressions for thirteen countries
and ®nds considerable differences in the intensity of the response of output
and prices to monetary shocks across countries; and Cecchetti and Rich
(1999), who look at a simple two-variable system for Australia, Canada, France,
Italy, Switzerland, the United Kingdom and the United States, and ®nd large
differences in the implied impacts.
Each of these studies has advantages and disadvantages. Overall, I have
chosen to examine the results reported by Ehrmann (1998). The appeal of
Ehrmann's approach is that it yields a series of estimates, all based on the same
methodology, for nearly the full set of EU countries. Ehrmann uses techniques
devised by King et al. (1991). In effect, Ehrmann identi®es monetary shocks
using a combination of long-run and short-run restrictions. The methods are
described both in his paper and in Cecchetti, McConnell and Perez Quiros
(1999). For each country, the model has either four or ®ve variables, including
output, in¯ation and an interest rate, and ± with the exception of Germany ±
an exchange rate. When a ®fth variable is present, it is either a second interest
rate or a commodity price index.16
The chart plots the responses of output and in¯ation to an interest rate
movement of 100 basis points for ten EU countries and the United States (see
Figure 5.1).17 These ten countries are the ones for which Ehrmann is able to
generate consistent and plausible results.18 As is clear from these plots, the
point estimates of the impulse response functions vary dramatically across
countries. Looking at the impact of interest rate movements on output, note
that for France and Germany, the peak impact is nearly twice what it is in the
remaining European countries, and ®fteen times the estimated impact in the
United States. The impact of policy on in¯ation also varies substantially.
Table 5.6 reports the maximum impact of a 100-basis-point monetary
contraction on output and in¯ation for all of the countries for which I have
estimates. I also include a measure of the timing of the impact ± the quarter at
which the maximum effect occurs. Column (5) presents a measure of the ratio
of the average output response to the average in¯ation response. This measure
is related to the sacri®ce ratio because it is roughly the output loss for an
in¯ation decline of one percentage point over a horizon of approximately
three years. Unfortunately, these estimates are not terribly precise, a point that
is clear from the results in Ehrmann's paper,19 and so we should not take some
of the numbers too seriously.
5 Systematic differences in national legal systems
If differences in ®nancial systems are creating the cross-sectional variation in

the transmission mechanism, it is natural to look for the causes of these
Table 5.6. Impact on output and in¯ation of a 100-basis-point increase in interest rates
184
Output In¯ation
Maximum Quarter of Maximum Quarter of Approximate

impact maximum impact impact maximum impact sacri®ce ratio
Country (1) (2) (3) (4) (5)

Austria Ð Ð Ð Ð Ð
Belgium ±0.72 2 ±0.05 1 ±45.29
Finland Ð Ð Ð Ð Ð
France ±1.30 5 ±0.21 2 ±12.07
Germany ±1.21 5 ±0.48 2 ±5.83
Ireland ±0.76 4 ±0.25 5 ±3.45
Italy ±0.64 5 ±0.25 9 ±5.01
Netherlands Ð Ð Ð Ð Ð
Portugal ±0.39 2 ±0.28 1 ±0.58
Spain ±0.46 4 ±0.23 4 ±1.34
Members of the EU cnot in EMU

Denmark ±0.48 5 ±0.34 1 ±1.69
Greece Ð Ð Ð Ð Ð
Sweden ±0.56 4 ±0.11 5 ±5.61
United Kingdom ±0.53 13 ±0.37 1 ±2.57
Other countries
Japan Ð Ð Ð Ð Ð
United States ±0.07 6 ±0.017 12 ±3.27
Sources: Estimates for the United States are from Cecchetti (1996); those for the remaining countries are from the estimation of Ehrmann's model in
Cecchetti, McConnell and Perez Quiros (1999).
differences. As noted earlier, La Porta et al. (1997, 1998) have examined the
relationship between a country's legal system and its ®nancial system. The
premise of their work is that investors provide capital to ®rms only if the
investors have the ability to get their money back. For equity holders, this
means that they must be able to vote out directors and managers who do not
pay them. For creditors, this means that they must have the authority to
repossess collateral. In addition to having nominal legal rights, these groups
must also have con®dence that the laws will be enforced.
La Porta et al. (1997, 1998) collect data on the legal systems in forty-nine
countries. They argue that all of these legal systems belong to one of four
families: English common law, French civil law, Scandinavian civil law and
German civil law. With regard to shareholder rights ± speci®cally, the ability of
shareholders to vote directors out ± English common law countries have the
best protections and French civil law countries have the worst. The pattern is
similar for creditor rights, which entail the right to reorganise or liquidate a
®rm. The pattern for enforcement is a bit different: Scandinavian civil law
countries have the most rigorous law enforcement, while French civil law
countries have the most lax.
Table 5.7. Shareholder rights, creditor rights and enforcement, byCountry
Shareholder rights Creditor rights Enforcement Legal family

Country (1) (2) (3) (4)

Austria 2 3 10.00 German
Belgium 0 2 10.00 French
Finland 2 1 10.00 Scandinavian
France 2 0 8.98 French
Germany 1 3 9.23 German
Ireland 3 1 7.80 English
Italy 0 2 8.33 French
Netherlands 2 2 10.00 French
Portugal 2 1 8.68 French
Spain 2 2 7.80 French

Denmark 3 3 10.00 Scandinavian
Greece 1 1 6.18 French
Sweden 2 2 10.00 Scandinavian
United Kingdom 4 4 8.57 English
Other countries
Japan 3 2 8.98 German
United States S 1 10.00 English
Source: La Porta et al. (1997), Table II.

Table 5.7 reproduces a portion of Table II from La Porta et al. (1997). The
column labelled `Shareholder rights' reports an index that is higher when
shareholders ®nd it less costly and dif®cult to vote directors out. The column
labelled `Creditor rights' reports an analogous index that is lower when
creditors experience less dif®culty gaining possession of property that has
been used to collateralise a bond or loan. Enforcement is an assessment of
countries' rigour in carrying out their laws, with a higher score implying more
aggressive enforcement. Finally, the table reports the legal family from which
each country's laws are derived.
Using these data to examine the relationship between shareholder rights
creditor rights, and enforcement on the one hand, and the concentration of
ownership and the availability of external ®nance on the other, La Porta et al.
(1997, 1998) come to two conclusions. First, corporate ownership is more
concentrated in countries where shareholders and creditors are poorly
protected by both the substance of the law and its enforcement. Second, and
more germane to the current discussion, countries with weaker legal rules and
less rigorous law enforcement have smaller and narrower capital markets.
Overall, English common law countries have the least concentration of
corporate ownership and the largest and deepest capital markets. French civil
law countries have the most concentrated ownership and the smallest capital
markets. La Porta et al. (1997 p. 1131) conclude that the `differences in the
nature and effectiveness of the ®nancial systems around the world can be
traced, in part, to differences in investor protection against expropriation by
insiders, as re¯ected by legal rules and the quality of their enforcement'. Their
®ndings are con®rmed by the data in Table 5.4, which show clearly that the
United States and the United Kingdom have much more extensive capital
markets than France and Italy.
6 Relationship of the legal environment to the impact of policy
Following the demonstration in La Porta et al. (1997, 1998) that the systematic
variation in systems of corporate governance and ®nance across countries can
be tied to the differences in the countries' legal systems, I ask if the variation in
the predicted strength of the lending channel and the estimated impact of
interest rate movements on output and in¯ation can be traced to these same
legal differences.20 To address this question, I combine the data from Table 5.5
on the predicted strength of the lending channel of monetary transmission
and from Table 5.6 on the size of the impact of interest rate movements
on output and in¯ation with the measures of cross-country differences in
legal organization from Table 5.7. In Table 5.8, I report the results of two
straightforward exercises. The ®rst separates the countries by the origin of their
legal system and constructs group averages for the effectiveness and impact of
monetary policy from column (4) of Table 5.5 and columns (1) and (3) of Table
5.6 (Table 5.8, top panel). The results follow the pattern predicted by the index
Table 5.8 Testing the relationship between cross-country differences in legal structure
and monetary policy effectiveness
Predicted Impact of policy

effectiveness of Approximate
Legal family monetary policy On output On in¯ation sacri®ce ratio
Group mean
English 1.1 ±0.45 ±0.21 ±3.1

Scandinavian 1.8 ±0.52 ±0.22 ±3.7
French 2.1 ±0.70 ±0.20 4.8a
German 2.4 ±1.25 ±0.49 ±5.8
Instrumental variables regression
Coef®cient Ð ±0.46 0.05 ±10.4

Standard error Ð (0.22) (0.08) (10.4)
Notes: `Predicted effectiveness' is drawn from column (4) of Table 5.5; the Ìmpact of policy', from
columns (1) and (3) of Table 5.6. The instrumental variables regression is of columns (1) and (3) of
Table 5.6 on column (4) of Table 5.5, with columns (1), (2) and (3) of Table 5.7 as instruments. All of
the results in this table use only the eleven countries for which there are estimates in Table 5.6:
Ireland, the United Kingdom and the United States (English common law); Denmark and Sweden
(Scandinavian common law); Belgium, France, Italy, Portugal and Spain (French civil law); and
Germany (German civil law).
a
average excludes Belgium.
of lending channel effectiveness as the impact of policy on output and the

approximate sacri®ce ratio vary systematically ± and as expected ± with the
origin of a country's legal system.
We can learn a bit more from the data than is recovered from the simple
averages reported in the top panel of Table 5.8. The question of greatest
interest is whether the cross-country heterogeneity in the real effects of
monetary policy can be explained by differences in the countries' ®nancial
systems, which have their source in the strength of shareholder and creditor
rights and the rigour with which these rights are enforced. We can do this
without fully accounting for all of the variation in the transmission
mechanism if we assume that the La Porta et al. (1997) measures are valid
instruments for the ®nancial variables in a simple regression that has the
impact of policy on the left-hand side and the overall measure of the lending
channel's effectiveness on the right-hand side. That is, I assume that the
shareholder, creditor and enforcement variables are exogenous, while the
measure of the effectiveness of the lending channel may not be.
The results of these two-stage least squares regressions are reported in the
bottom panel of Table 5.8. Again, we see that the countries in which the
lending channel is expected to be strongest have the biggest sacri®ce ratios
and show the largest impact of interest rate movements on output. The latter
of these relationships has a t-ratio of 2.1, and so it may even be signi®cantly
different from zero. The results for in¯ation are much less satisfactory: the
measures of ®nancial structure appear to be uncorrelated with the impact of
policy on prices. Because of the small size of the sample (eleven countries), the
estimates are all fairly imprecise, and so I treat them as being only suggestive.
7 Concluding remarks
Among the many challenges facing the new Eurosystem is the possibility that
the regions of the Euro Area will respond differently to interest rate changes. In
this chapter, I have suggested that differences in ®nancial structure are a
proximate cause for these national asymmetries in the monetary policy
transmission mechanism. Moreover, I have proposed that these differences in
®nancial structure are likely a result of the EU countries' diverse legal
structures. The evidence, although circumstantial, is consistent with this view.
Most economists believe that the monetary transmission mechanism will vary
systematically across countries with differences in the size, concentration and
health of the banking system, and with differences in the availability of
primary capital market ®nancing. The countries of the EU differ quite
dramatically in all of the dimensions that would seem to matter, leading to the
prediction that the impact of interest rates on output and prices will not be
consistent across countries. While the estimates of the impact of interest rate
changes on output and in¯ation tend to be quite imprecise, they do differ, and
in the way that is predicted by the state of the countries' ®nancial systems.
Finally, we can trace differences in ®nancial structure, the size and scope of
capital markets, and the availability of alternatives to bank ®nancing to
differences in the countries' legal structures.
What does this mean for the future of ®nancial markets and monetary
policy in the Euro Area? Will the European banking system become more like
that of the United States? The arguments presented here suggest that unless
legal structures are harmonised across Europe, ®nancial structures will remain
diverse, and so will monetary transmission mechanisms. It will not be enough
to make regulatory structures more similar, since such a change will not, in
and of itself, alter the structure of capital markets. In other words, I do not view
regulatory competition as a force to eliminate the asymmetries in the ®nancial
intermediation systems of the EU.21 As the ECB (1999) report makes clear, this
force has been very weak in the past and is expected to be weak in the future.
While we may see cross-border mergers and acquisitions of ®nancial sector
®rms that take advantage of the expertise of those already doing business in a
region,22 only a decision to change the existing legal structures so that
shareholders and creditors in all EU countries enjoy the same rights will force
the movement to a US- style ®nancial structure.
Data Appendix
The data sources for Tables 5.2±5.4 in this chapter are as follows:
Table 5.2
Number of institutions and concentration ratios: For all countries, concentration is
calculated as the assets of the top ®ve banks as a percentage of total bank assets.
Population: International Monetary Fund, International Financial Statistics (January
1999), country report tables, l. 99z, mid-year estimates for all countries.
Austria: Austrian National Bank web pages < http://www.oenb.co.at/stat-monatsheft/
tabellen/2001p.htm >, Ingesamt, Hauptanstalten, for number of institutions; and
< http://www.oenb.co.at/stat-monatsheft/tabellen 2000_5p.htm >, Alle Sektoren, Summe
Aktiva (Ohne Rediskonte), for total assets; Austrian National Bank, Economic Analysis
Division, for assets of top ®ve and top ten banks.
Belgium: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports
on bank balance sheets, p. 36, l. 37 (under supplementary information), for number of
institutions; Bank of Belgium, Financial and Economic Statistics Division, for total assets
of credit institutions and for share of top ®ve banks.
Finland: Bank of Finland, Financial Statistics Desk, for all ®gures.
France: Bank of France, Monetary Research and Statistics Division (DESM-SASM) for all
®gures on credit institutions.
Germany: Deutsche Bundesbank, Monthly Report (May 1998), p. 16, Table IV.1, column
1, for number of institutions; Deutsche Bundesbank, Department of Controlling,
Accounting and Organisation, Division C-2, for share of top ®ve banks.
Ireland: Central Bank of Ireland, Monetary Policy and Statistics, for number and total
assets of all credit institutions (which include licensed banks, building societies, state-
sponsored ®nancial institutions and savings banks); IBCA BankScope database, for assets
of top ®ve banks.
Italy: Bank of Italy, Research Department, for all ®gures.
Netherlands: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, p. 192, l. 37 (under supplementary information), for
number of institutions; De Nederlandsche Bank, Annual Report (1997), Tables 1, 2.1 and
2.2, for assets of top ®ve banks and for total assets of monetary institutions.
Portugal: Bank of Portugal web page < http://www.bportugal.pt/publish/frpu-
blish_e.htm >, Chart VIII.1 and Table VIII.2, for number of institutions and share of
top ®ve banks. OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, p. 231, l. 25, for total assets of commercial banks.
Spain: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country reports on
bank balance sheets, p. 236, l. 37 (under supplementary information), for number of
banks; Bank of Spain, Statistical Bulletin (June 1998), Tables 61.1 (p. 271), 62.1 (p. 281),
63.1 (p. 291), sum of column 1 in all tables, for total assets of banks, savings banks, and
credit cooperatives; IBCA Bankscope database, for assets of top ®ve banks.
Denmark: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, p. 64, l. 37 (under supplementary information), for
number of institutions; Denmark National Bank web page < http://www.nationalban-
ken.dk/nb/nb.nsf/alldocs/ F15D9E8CF275ED1A2412565B4003E8BD5, > for total assets;
IBCA BankScope database, for assets of top ®ve banks.
Greece: Hellenic Bank Association, The Greek Banking System (April 1998), p. 87, for
number of institutions, total assets, and assets of top ®ve banks.
Sweden: Sveriges Riksbank, Statistical Yearbook (1996), p. 17, Table 6, for number of
banks; Sveriges Riksbank, Financial Statistics Department, for share of top ®ve banks.
United Kingdom: British Bankers Association, Annual Abstract of Banking Statistics
(1997), Table 1.04, for number of institutions; Bank of England, MFSD, for shares of top
®ve banks (data relate to all banks and building societies operating in the United
Kingdom and so include the business of foreign-owned af®liates in the United
Kingdom).
Japan: Bank of Japan, International Department, for all ®gures for banks and other
deposit-taking institutions, end of ®scal year 1996 (March 1997).
United States: Federal Financial Institutions Examination Council (FFIEC), Reports of
Condition (call reports database), for all ®gures for commercial banks.
Table 5.3
Bank data: McCauley and White (1997), Table 1. Federal Reserve Bank of New York staff
calculations for Austria, Belgium, Greece, Ireland, Italy and Portugal, based on ranking of
asset size from IBCA BankScope database. In each country, banks were chosen according
to 1997 assets. Return on assets, loan loss provisions, net interest margin and operating
cost are drawn from IBCA BankScope database.
Thomson ratings: Thomson BankWatch database.
Table 5.4
Number of publicly traded ®rms and market capitalisation: International Finance
Corporation, Emerging Stock Markets Factbook (1997), pp. 17 and 23 (also available on
the Wall Street Journal web site < http://update.wsj.com/public/resources/documents/gi-
tab5.htm >).
Population: See sources for population data in Table 5.2.
Privately issued debt: Bank for International Settlements, International Banking and
Financial Market Developments (February 1998), pp. 46±7, Tables 14 and 15, amount
outstanding, December 1996 ®gures; sum of ®gures from Table 14 (international debt
securities) and Table 15 (domestic debt securities).
GDP: International Monetary Fund, International Financial Statistics (January 1999),
country report tables, l. 99b.c for all countries. Year-average exchange rates used for
conversion into US dollars (local currency per US dollar, l. rf for all countries).
Bank loans: OECD, Bank Pro®tability: Financial Statements of Banks (1998), country
reports on bank balance sheets, l. 16 on pp. 27, 35, 63, 67, 91, 115, 143, 159, 163, 167,
191, 231, 235, 251, 259, 263, 303, 307 and 315.
Notes
1. I gratefully acknowledge comments from and discussions with Ignazio Angeloni,
Paul Bennett, Stefan Gerlach, Heinz Hermann, Beverly Hirtle, James Kahn, Anil
Kashyap, Kenneth Kuttner, Gabriel Perez Quiros, Margaret Mary McConnell,
Robert Rich and especially Paolo Pesenti. I thank Michael Ehrmann for providing
his programs, and Rama Seth for helping with the data.
2. Throughout the chapter, I refer to the eleven countries of EMU but provide
information on only ten. Luxembourg is not included.
3. See Angeloni and Dedola (1998).
4. Similar points are made by White (1998), who suggests that competition in
banking may be about to increase in Europe, stimulated by the introduction of the
Euro. In addition, a ECB (1999) study suggests that EMU may speed up the process
of disintermediation and lead to a more geographically diversi®ed and inter-
nationalised banking system.
5. For example, within the United States, more than 10 per cent of ®rms with assets
exceeding $1 million have chosen to incorporate in Delaware, a state with less than
half of one per cent of the country's population. Why is this? The answer can be
found by considering how the development of Delaware's legal structure has
differed from the development of the legal structure in other states. Originally,
large ®rms were incorporated in New Jersey because the state, in exchange for
incorporation fees and franchise taxes, had liberalised its corporation law to allow
various mergers and cross-holdings that were disallowed elsewhere. State law also
gave very strong power to corporations' directors (Grandy, 1989). Delaware copied
New Jersey's statutes and then bene®ted from changes made to New Jersey's law by
Governor Woodrow Wilson in 1913. As this example suggests, the economic
structure has its source in the legal structure.
6. I should note that ®rms in countries that act slowly will be put at a competitive
disadvantage, and so they might pressure their governments to speed up the
legal changes. The potential strength of such regulatory competition is an open
issue.
7. There is an alternative. A company may move to a country where the ®nancial
system better suits its needs. The La Porta et al. measures, reported in Table 5.7,
suggest that the United Kingdom is the best country in the European Community
in which to issue both bonds and stocks, and so ®rms that wish to have ready
access to primary capital market ®nancing may tend to concentrate there. But for
this strategy to be successful, ®rms would have to reincorporate and move assets
into the alternative jurisdiction. The assets must move to provide the proper
guarantees to investors. All of this seems unlikely.
8. In addition to the differences in the type of nominal rigidity, there are variations in
the way in which the rigidites are modelled. These variations are more than formal;
they have very different implications for the dynamic effects of nominal shocks on
real variables. Different modelling strategies are based on differences in the timing
of price or wage change decisions. There are three basic schemes used, based on
Fischer (1977), Taylor (1980) and Calvo (1983), and they create very different
dynamic responses of real variables to nominal shocks. Fischer, for example,
assumes prices are predetermined, meaning that at some time agents set prices for
some number of future periods: the level of prices set on the decision date can differ
for the different periods before the next decision date. In this model, the impact of
a nominal shock lasts for only as long as it takes for all price setters to have a chance
to reset their price schedules. In the Taylor model, prices or wages are assumed to be
®xed, meaning that their nominal value does not vary between decision dates.
When prices or wages are ®xed, nominal shocks die out only asymptotically. In
Calvo's (1983) model, price-setters change their prices according to a Poisson
process, leading to a variety of possible dynamics.
9. Longer-run considerations, such as the potential costs or bene®ts of modest levels
of in¯ation, critically depend on understanding the sources of nominal rigidity. For
example, Akerlof, Dickens and Perry (1996) and Groshen and Schweitzer (1997)
consider whether small positive levels of aggregate in¯ation can facilitate real
adjustments in the presence of an aversion to nominal wage declines, suggesting
that the long-run goal for in¯ation might be positive. But Feldstein (1996)
contends that the tax distortions created by in¯ation reduce the level of output
permanently, an argument that suggests that the optimal level of in¯ation may
even be negative. Overall, most economists now seem to agree that in¯ation leads
to lower levels of real output and may even retard long-run growth. See Feldstein
(1999) for a summary.
10. As emphasised by Kashyap and Stein (1994), this assertion applies to both ®nancial
and non-®nancial ®rms.
11. This is not to say that the traditional mechanisms, operating through interest rates
and exchange rates, are not present as well. Unfortunately, it has proved to be very
dif®cult to disentangle the individual importance of the various channels of
transmission.
12. It is important to note that there can be signi®cant cyclical and secular changes in
the strength of the lending channel as the health and concentration of the banking
system change, and as capital markets become deeper and broader.
13. After I collected the data for this section, the ECB issued its report Possible Effects of
EMU on the EU Banking Systems in the Medium to Long Term. The Appendix tables in
that report contain much of the same information presented here.
14. Throughout the analysis, I omit Luxembourg.
15. A signi®cant failing of this analysis is the assumption that these relative rankings
are not changing over time. Surely, if I had chosen different dates to measure the
relative health and concentration of countries' banking systems, I would have
created a different set of rankings for the ®rst two indicators. It is entirely possible
that both the relative importance of small banks and the health of the banking
system will become increasingly uniform across countries, leaving only differences
in external ®nance.
16. See Appendix A in Ehrmann (1998) for additional details.
17. The results for the United States are derived from Cecchetti (1996).
18. Although he reports estimates for thirteen countries, the estimates for three of
these countries appear to be dif®cult to interpret. In the case of Finland, for
example, the impact of monetary tightening is to increase output, not decrease it.
For Austria and the Netherlands, we have not been able to replicate the results in
the current version of Ehrmann's paper.
19. Figures 1±13 in Ehrmann (1998) show that the impulse response functions are
rarely signi®cantly different from zero. The same point is made in Cecchetti (1998)
and Cecchetti and Rich (1999).
20 White (1998) makes a related point when he notes that the legal, tax, regulatory
and supervisory frameworks within which ®nancial institutions operate differ
signi®cantly across the various countries of the EU. All of these differences make
direct competition more complex and less appealing. He goes on to focus on
differences in the EU countries' labour laws and in the regulatory restrictions the
countries place on the types of ®nancial products that can be offered. These effects
are surely complementary to the ones I address here.
21. It is also extremely unlikely that these dif®culties will be overcome by the
issuance of debt and equity in a jurisdiction that offers suf®cient investor
protections. But unless ®rms have assets within these jurisdictions, I do not see
this as a solution.
22. Such developments would be similar to what has happened with the relaxation of
interstate branching regulations in the United States, where banks in one state have
purchased a bank in another state in order to obtain the legal and regulatory
knowledge to do business in that state. Interstate branching has not meant
opening new branches of an existing bank in another region.
References
Akerlof, G. A., W. T. Dickens and G. L. Perry, (1996) `The Macroeconomics of Low
In¯ation', Brookings Papers on Economic Activity, 1, 1±59.
Angeloni, I., and L. Dedola, (1998) `From the ERM to the Euro: A Soft Transition?', Banca
d'Italia, Working paper, June.
Bernanke, B. S. (1983) `Nonmonetary Effects of the Financial Crisis in the Propagation of
the Great Depression', American Economic Review, 73, 257±76.
ÐÐÐÐ (1993) `Credit in the Macroeconomy', Federal Reserve Bank of New York Quarterly
Review, Spring, 50±70.
Monetary Transmission', American Economic Review, 82, 901±21.
Bernanke, B. S. and M. Gertler (1989) Àgency Costs, Net Worth and Business
Fluctuations', American Economic Review, 79, 14±31.
ÐÐÐÐ (1990) `Financial Fragility and Economic Performance', Quarterly Journal of
Economics, 105, 87±114.
Calvo, G. A. (1983) `Staggered Price Setting in a Utility-Maximizing Framework', Journal
of Monetary Economics, 12, 383±98.
Cecchetti, S. G. (1995) `Distinguishing Theories of the Monetary Transmission Mechanism',
Federal Reserve Bank of St. Louis Economic Review, 77, 83±97.
ÐÐÐÐ (1996) `Practical Issues in Monetary Policy Targeting', Federal Reserve Bank of
Cleveland Economic Review, 32, 2±15.
ÐÐÐÐ (1998) `Policy Rules and Targets: Framing the Central Banker's Problem', Federal
Reserve Bank of New York Economic Policy Review, 4, 1±14.
Cecchetti, S. G., M. M. McConnell and G. Perez Quiros (1999) `Policy-makers' Revealed
Preferences and the Output±In¯ation Variability Trade-off', Federal Reserve Bank of
New York, February, unpublished.
Cecchetti, S. G. and R. W. Rich (1999) Ìnternational Evidence on the Sacri®ce Ratio:
Do We Know How Big It Is?', Federal Reserve Bank of New York, February,
unpublished.
Christiano, L. J., M. Eichenbaum and C. L. Evans (1997) `Sticky Price and Limited
Participation Models of Money: A Comparison', European Economic Review, 41,
1201±49.
Danthine, J.-P., F. Giavazzi, X. Vives and E.-L. von Thadden (1999) The Future of European
Banking, London, Centre for Economic Policy Research.
de Bondt, G. J. (1997) `Monetary Transmission in Six EU-Countries: An Introduction and
Overview', De Nederlandsche Bank, Research Memorandum 527. Dornbusch, R.,
C. Favero and F. Giavazzi (1998) Ìmmediate Challenges for the European Central
Bank', Economic Policy, 28, 17±64.
Ehrmann, M. (1998) `Will EMU Generate Asymmetry? Comparing Monetary Policy
Transmission Across European Countries', European University Institute Working Paper
98/28, October.
European Central Bank (1999) Possible Effects of EMU on the EU Banking Systems in the
Medium to Long Term, Frankfurt, European Central Bank, February.
Feldstein, M. (1996) `The Costs and Bene®ts of Going from Low In¯ation to Price
Stability', In C. Romer and D. Romer (eds)., Reducing In¯ation, Chicago, University of
Chicago Press, 123±66.
Feldstein, M. (ed.) (1999) The Costs and Bene®ts of Achieving Price Stability, Chicago,
University of Chicago Press for NBER.
Fischer, S. (1977) `Long-term Contracts, Rational Expectations, and the Optimal Money
Supply Rule', Journal of Political Economy, 85, 191±205.
Gerlach, S. and F. Smets (1995) `The Monetary Transmission Mechanism: Evidence from
the G-7 Countries', Centre for Economic Policy Research Working Paper 1219, July.
Gertler, M. and S. Gilchrist (1993) `The Role of Credit Market Imperfections in the
Monetary Transmission Mechanism: Arguments and Evidence', Scandinavian Journal
of Economics, 95, 43±64.
Grandy, C. (1989) `New Jersey Corporate Chartermongering, 1875±1929', Journal of
Economic History, 49, 677±92.
Groshen, E. and M. Schweitzer (1997) Ìdentifying In¯ation's Grease and Sand Effects in
the Labor Market', NBER Working Paper 6061, June.
Hubbard, R. G. (1995) Ìs There a `Credit Channel' of Monetary Policy?', Federal Reserve
Bank of St Louis Economic Review, 77, May-June, 63±77.
Kashyap, A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending', In N. G.
Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press for NBER.
ÐÐÐÐ (1997) `The Role of Banks in Monetary Policy: A Survey with Implications for
the European Monetary Union', Federal Reserve Bank of Chicago Economic Perspectives,
September±October, 2±18.
Kieler, M. and T. Saarenheimo (1998) `Differences in Monetary Policy Transmission? A
Case Not Closed', European Commission Economic Papers, 132, November.
King, R. G., C. I. Plosser, J. H. Stock and M. W. Watson (1991) `Stochastic Trends and
Economic Fluctuations', American Economic Review, 81, 819±40.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny (1997) `Legal Determinants
of External Finance', Journal of Finance, 52, 1131±50.
ÐÐÐÐ (1998) `Law and Finance', Journal of Political Economy, 106, 1113±55.
McCauley, R. N. and W. R. White (1997) `The Euro and European Financial Markets',
Bank for International Settlements Working Paper 41, May.
Rotemberg, J. J. (1984) À Monetary Equilibrium Model with Transactions Costs', Journal
of Political Economy, 92, 40±58.
Taylor, J. B. (1980) Àggregate Dynamics and Staggered Contracts', Journal of Political
Economy, 88, 1±23.
Vlaar, P. J. G. and H. Schuberth (1998) `Monetary Transmission and Controllability of
Money in Europe: A Structural Vector Error Correction Approach', De Nederlandsche
Bank, May, unpublished.
White, W. R. (1998) `The Coming Transformation of Continental Banking?', Bank for
International Settlements Working Paper 54, June.
Discussion
Manfred J. M. Neumann
Cecchetti provides us in Chapter 5 with an empirical study of the monetary

transmission mechanism in the European countries. Studying country-speci®c
differences of the transmission process is potentially of great relevance in view
of the fact that European Monetary Union (EMU) is in operation. Now the
eleven countries that have entered EMU are confronted with the single
monetary policy run by the independent European Central Bank (ECB). Given
the diversity of economic and ®nancial structure it is likely that the ECB's
monetary policy actions will affect the member economies at different
strengths and speeds.
The focus of Cecchetti's study is on assessing the empirical relevance of the
bank credit channel of policy transmission which forms an important part of
the transmission mechanism of relative prices. The theory of relative prices
states that any policy shock destroys portfolio equilibrium and affects all
®nancial and real assets and, consequently, the net worth of ®nancial
intermediaries, enterprises and households. Owing to asset-speci®c informa-
tion and transaction costs as well as rigidities, the shock transmission is not
uniform but differs among assets as regards price and quantity response as well
as timing. The credit view concentrates on the differential impact of policy
shocks on the supply of loans by banks versus the supply of credit by other
®nancial intermediaries or markets. Given that banking regulation makes
banks dependent on the supply of reserve money, the prediction is that
monetary policy affects banks, hence their clients, quicker and more forcefully
than credit markets outside the banking system.
Cecchetti's study essentially consists of three parts. Following the work by
Kashyap and Stein (1997), he collects data on selected characteristics of the
®nancial structure of fourteen EU countries as well as the United States and
Japan. For each country the information is condensed in two steps: ®rst, three
partial indicators are computed that are supposed to re¯ect the relative
importance of small banks, the health of the banking system and the
availability of alternative ®nance; second, a summary indicator of the strength
of the bank credit channel is computed, de®ned as the arithmetic mean of the
195
196 Manfred J. M. Neumann
three partial indicators. Cecchetti then addresses the question whether the
evidence on country differences in credit channel strength can be explained
by differences in the countries` legal systems as regards civil law. Finally ± and
this an innovative aspect of the study ± Cecchetti examines whether the
computed summary index of credit channel strength or importance is
correlated with separate evidence, derived from estimates of structural VARs
on the responses of output and in¯ation to monetary policy shocks. He ®nds
that the expected correlation exists.
While Cecchetti's research agenda is a potentially very fruitful one, the
present study is a ®rst, very tentative exploration. My reading of the evidence
presented is that it does not lend support to the prediction that the impact of
monetary policy is signi®cantly shaped by differences in countries' ®nancial
structures.
1 Data evaluation
Cecchetti presents data that describe ®nancial sector characteristics of sixteen

countries and aggregates them to derive three indicators that are intended to
re¯ect the relative strength of the credit channel. A fundamental problem to
be noted is that the indicators are not strictly based on some aggregation
formula but are affected by subjective evaluation. Cecchetti's indicator of the
importance of small banks can be used to make the point. In the following
Table D5.1 I show the two characteristics Cecchetti apparently relies on:
(1) the number of banks per million people, and (2) a concentration ratio
that aggregates the shares of the top ®ve banks in total bank assets.
Compare, for example, Greece with the United Kingdom. Greece is assigned an
indicator value of 2 while the United Kingdom receives a 1, meaning that small
banks are less important in the United Kingdom than in Greece. But in fact,
Greece has less banks per million of people than the United Kingdom (two to be
compared to eight) and the Greek concentration ratio is three times the British
ratio. Moreover, as the last column of Table D5.1 indicates, the average asset share
of a Greek bank (excluding the top ®ve) is twelve times the share a British bank
enjoys. On all accounts, it seems that the United Kingdom should be assigned an
indicator value of 2 instead of 1 and Greece a value of 1 instead of 2.
As a second example, compare Austria with the United Kingdom.
Cecchetti's indicator signals that in Austria small banks are more important
than in the United Kingdom. Indeed, on account of the number of banks per
million of people this judgement is justi®ed. Looking at the concentration
ratio, in contrast, the credit channel is less strong in Austria than in the United
Kingdom. The top ®ve Austrian banks control 48 per cent of the banking
system's assets while in the United Kingdom the top ®ve control only 28 per
cent. Thus, we are unable to predict from these data whether a monetary
policy shock will have a larger impact in Austria or in the United Kingdom.
Given this cognitive problem, Cecchetti resorts to subjective weighting as
Discussion 197
Table D5.1 On the importance of small banks
For
Banks Concentration Indicator comparison:
per million ratio: of the average
people top ®ve banks importance asset share,
of excluding
small banks top ®ve
Germany 43 17 3 0.02
USA 40 17 3 0.01
Austria 126 48 3 0.05
Denmark 22 17 2 0.74
Italy 16 25 2 0.08
Japan 4 30 2 0.13
France 24 40 2 0.04
Spain 8 44 2 0.18
Greece 2 71 2 1.93
United Kingdom 8 28 1 0.15

Belgium 14 57 1 0.32
Ireland 18 41 1 1.04
Finland 68 78 1 0.06
Netherlands 11 79 1 0.13
Portugal 5 76 1 0.52
Sweden 14 90 1 0.08
Sources: Cecchetti (this volume), Tables 5.2 and 5.5; own computations.
Kashyap and Stein (1997) have done. But this practice throws considerable
doubt on using this as well as similar indicators as inputs in subsequent
empirical work.
The more types of characteristics are considered, the more relevant becomes
the danger of biased aggregation. Table D5.2 shows the two characteristics
used by Cecchetti to derive an indicator of the availability of alternative
®nance. Countries with the least (most) developed external capital markets
receive an indicator value of 3 (1). Compare, for example, Finland with
Belgium. To be sure, on both characteristics the credit channel should be
stronger in Belgium than in Finland, implying that the rank order should be
reversed. Alternatively, compare Greece with Germany. The ratio of market
capitalisation is smaller in Greece but so is the share of bank loans in total
®nance. Thus, it is not clear on what grounds Cecchetti assumes that the credit
channel is stronger in Greece than in Germany.
Given these inconsistencies in the data transformation, the information
content of Cecchetti's indicators is in doubt. I realise that the researchers of the
credit channel are aware of the crude nature of this work, and I admit that I do
not know how to overcome the methodological problem of aggregating
Table D5.2 Availability of alternative ®nance
Market Bank loans Indicator

capitalization as a percentage of of
as a percentage all forms of availability
of GDP ®nance of alternative
®nance
Austria 15 65 3
Ireland 18 80 3
Greece 20 48 3
Italy 21 50 3
Portugal 23 62 3
Germany 29 55 2
France 38 49 2
Spain 42 58 2
Finland 50 39 2
Japan 67 59 2
Netherlands 96 53 2
Denmark 41 25 1
Belgium 45 49 1
Sweden 99 32 1
United Kingdom 150 37 1
United States 111 21 1
Source: Cecchetti (this volume), Tables 5.4 and 5.5.
con¯icting information. Nevertheless it seems that the subjective weighting

procedure should be abandoned.
2 Is the evidence in favour of the credit channel?
A novel feature of Cecchetti's study is that he computes a summary indicator

of credit channel strength and confronts the predictions embodied in this
indicator with estimates of the impact of monetary policy on output and
in¯ation. The estimates are taken from a study by Ehrmann (1998) that
employs structural VARs. This information, too, is aggregated for different
groups of countries. The guiding notion for the grouping is the nature of the
countries' legal system as regards civil law. La Porta et al. (1997) pointed out
that the legal protection of creditors and shareholders is similar across
countries that share the same legal tradition but very different across legal
traditions.
By comparing group averages Cecchetti shows in his Table 5.1 (p. 173) as
well as the upper panel of his Table 5.8 (p. 187) that there apparently is a
positive correlation between the estimated (maximum) impact of monetary
Discussion 199
policy on output (though not on in¯ation) and the prediction of relative credit
channel strength derived from his summary indicator of ®nancial structures.
As an additional underpinning Cecchetti also provides a supporting two-stage
least squares regression that employs legal information as instrumental
variables. If valid, this is a nice result. Unfortunately, there is reason to
conjecture that the result is an artefact.
Cecchetti himself notes that the VAR estimates are not very precise,
meaning that the con®dence bands for most countries are very large (see
Ehrman, 1998). Moreover, the estimates suggest for six out of eleven countries
that the maximum impact of monetary policy on in¯ation precedes the
impact on output, in most cases by several quarters. To be sure, this timing is
dif®cult to rationalise and hence throws doubt on the reliability of the
estimates. But let us assume that the VAR estimates for output re¯ect the true
impact of monetary policy. On this assumption Cecchetti's regression appears
to support the conjecture that monetary policy impacts on output are stronger
the stronger is the credit channel. To check for this, Figure D5.1 plots the
output responses against the summary indicator of the credit channel. The
®gure reveals that France and Germany, and possibly the United States, are
0.0
USA
DK P
–0.5 E
Output response
UK S
I
B
IRE
–1.0
G
F
–1.5
0.5 1.0 1.5 2.0 2.5 3.0
Summary indicator of credit

channel strength
Figure D5.1 Maximum impact of an interest rate shock and strength of the credit
channel
outliers. For the remaining eight countries, who belong to different legal
traditions, we observe about the same output response to monetary policy.
In sum, we have to conclude that the evidence collected by Cecchetti does
not support the hypothesis that the effectiveness of monetary policy is
affected by a country's legal tradition or ®nancial structure. However, I hasten
to add that this is not a ®nal verdict on the role of the credit channel as the
negative result might be shaped by the shaky nature of the data employed.
References
Cecchetti, S. G. (2000) `Legal Structure, Financial Structure, and the Monetary Policy
Transmission Mechanism', Chapter 5 in this volume.
Ehrmann, M. (1998) `Will EMU Generate Asymmetry? Comparing Monetary Policy
Transmission Across European Countries', European University Institute Working Paper
98/28, October.
Kashyap, A. K. and J. C. Stein (1997) `The Role of Banks in Monetary Policy: A Survey
with Implications for the European Monetary Union, Federal Reserve Bank of Chicago
Economic Perspectives, September±October, 2±18.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny (1997) `Legal Determinants
of External Finance', Journal of Finance, 52, 1131±50.
Discussion1
Ignazio Angeloni
This is a very interesting and stimulating chapter, one that offers a novel
perspective on the transmission process of monetary policy. As in many of his
previous contributions, Cecchetti is able to identify a relevant policy issue,
bring together stimulating evidence and draw intuitively appealing inter-
pretations from it. If the prime test of the success of a study is to stimulate
thinking, this is unquestionably a successful study. No matter how much we
may disagree with the interpretations it proposes and with the overall policy
message ± and, as I will argue below, there is ample room for disagreement ± I
believe that future research on monetary transmision issues, related to the
Euro Area in particular, will have to take the ideas that Cecchetti presents here
carefully into account.
In my comments I will ®rst explain what, in my view, is the main
contribution of this chapter and outline the essence of the argument; I will
then focus on what I regard as limitations of the analysis and explain why and
how these limitations affect the overall policy conclusions.
The chapter builds an ambitious bridge between two so far separate strands
of literature ± that on the bank credit channel of monetary policy transmission
and that on the legal determinants of corporate ®nance and of ®nancial
structure in general. Popularised in the 1990s by contributions by Kashyap
and Stein (1994, 1997), Hubbard (1995), Gertler and Gilchrist (1993), building
on older ideas,2 the `lending view' predicts that the effects of monetary policy
on aggregate demand, output and in¯ation should be in¯uenced (normally
ampli®ed, as shown by Bernanke and Blinder, 1992) by the autonomous
behaviour of commercial banks. The key condition for this to happen is that
bank loans be imperfectly substitutable for other assets and liabilities, both in
the balance sheet of commercial banks and among the forms of ®nancing
available to ®rms and households. The monetary transmission process should
thus be related to the structural (legal, regulatory, cultural, etc.) features of
each country's banking and ®nancial sector that affect the extent to which
direct bank lending can be substituted by other types of credit. As this chapter
convincingly argues, a natural connection arises here with the so-called `legal
201
202 Ignazio Angeloni
view' of corporate ®nance and ®nancial structures, put forward more recently by
La Porta et al. (1997, 1998) and Rajan and Zingales (1998). According to this
literature, cross-country differences in corporate ®nancial structures are related to
the way in which national legal systems accord protection to different classes on
investors. Speci®cally, the `legal view' predicts that better creditor protection will
facilitate external ®nance and help the development of active capital markets, at
the disadvantage of bank intermediation, and that shareholder protection
should make equity ownership more attractive to investors and thus favour
equity relative to debt as a source of capital for non®nancial ®rms.
Having laid down this conceptual premise, the chapter then applies it to the
present situation of the Euro Area. This application is indeed tempting because
the eleven European countries that have adopted a common currency still
display considerable differences in ®nancial structures, banking regulatory
and supervisory frameworks, let alone the legal systems in a broader sense.
Consequently, the chapter's argument goes, the transmission of the single
monetary policy to the Euro Area economy cannot be but diverse across
countries, and will continue to be so unless and until radical steps are taken to
create a common legal framework in the area.
Unfortunately ± or perhaps one should say fortunately for the European Central
Bank (ECB) and the success of its monetary policy ± after closer inspection such
strong policy implication appears less convincing than the basic premise on
which it is built. This is not only because, as one concludes from Table 5.7
(p. 185), legal systems in the Euro Area are not so diverse after all ± the only
country with an Anglo±Saxon legal system is Ireland, and those belonging to the
German or French systems, whose implications for monetary policy effective-
ness are broadly similar, cover nine out of eleven countries or 97 per cent of the
area's GDP ± but for other and more substantive reasons as well. I will organise
my comments on them under three headings: (1) the role of other channels of
monetary transmission (speci®cally, the exchange and interest rate channels);
(2) the changing nature of the bank lending channel in the Euro Area; (3) the
robustness (or lack thereof) of some of the chapter's empirical results.
1 Exchange and interest rate effects
The analysis of this chapter is limited exclusively to the role of commercial

banks in the transmission process. However, a more complete and balanced
view of the effects of Stage Three of EMU should consider also the changes that
will take place in other channels of monetary transmission, particularly the
exchange rate and the (structure of) interest rates. From the viewpoint of these
two channels, the adoption of a single currency in the Euro zone should
unambiguously result in a more homogeneous transmission of monetary
policy across countries.
Starting with the exchange rate, let us distinguish its contribution to the
transmission of monetary policy in two steps. First, the exchange rate reacts to
Discussion 203
changes in the central bank controlled interest rates. This stage of the
transmission chain, which crucially depends on monetary policy credibility
and on the ®scal±monetary policy mix, is obviously identical across countries
in a monetary union, while it could differ sharply when each participating
central bank was acting in isolation within its own institutional and national
macroeconomic policy framework. The second step, the impact of exchange
rate changes on the economy, depends on the economy's degree of openness
to international trade. Table D5.3 shows the percentage ratios exports over
GDP in the individual countries of the Euro Area and in the area as a whole,
where in the latter case only extra-area trade is considered. Not only is the
average degree of openness of the area as a whole lower than that of each of the
participating countries, but its dispersion across countries is also signi®cantly
reduced when attention is focused on extra-area trade only. The impact on
individual countries of any given change of the exchange rate of the Euro is
less differentiated than those of each country's exchange rate changes before
monetary union. This argument can be extended to the role that yield curve
changes play in the transmission mechanism. Again, two steps can be
distinguished. First, monetary policy shocks affect the level and shape of the
yield curve, through a mix of liquidity and expectations effects. It has been
shown (Angeloni and Rovelli, 1998) that the impact on the yield curve of
monetary policy changes varies across monetary and ®scal regimes, and that it
is related to monetary policy credibility. Such differences, again, are
eliminated in a monetary union, where a unique yield curve prevails across
the whole area with the only exception of small cross-country spreads that
depend on bond market liquidity and on different credit ratings of sovereign
debtors. The second step is given by the reaction of the private sector's
spending behaviour to yield curve changes. Here the judgement of the
possible effects of the adoption of the single currency is, as in the case of the
exchange rate, more complex. Interest rate effects depend partly on the shape
of private agents' intertemporal preferences, for which no change should
a priori be expected. In addition, life cycle theory suggests also that these
effects should depend also on the size and composition of the household
sector's ®nancial wealth. The income effects of a given interest rate change will
be larger, ceteris paribus, the shorter the ®nancial duration of the household net
®nancial wealth and, for a given duration, the higher the size of wealth. Euro
Area countries have historically been characterised by widely divergent
degrees of liquidity of ®nancial assets, due to different public debt manage-
ment policies.3 As a result of EMU, and in the context of lower and more stable
in¯ation, a convergence process is taking place across Euro Area countries in
the maturity structure of public debts; moreover a convergence of debt ratios
will result from the application of the Stability and Growth Pact. As this
process unfolds, income and substitution effects on consumers following
interest rate changes are likely to follow a more homogeneous pattern within
the Euro Area.
Table D5.3 Openness to international trade in the Euro Area
Total exports/ Share outside Exports outside

GDP EU-11 the Euro Area/
GDP
Belgium±Luxembourg 62.6 42.8 26.8

Germany 24.3 57.7 14.0
Spain 19.6 41.2 8.1
France 20.4 51.6 10.5
Italy 20.8 56.2 11.7
Netherlands 43.4 42.7 18.5
Austria 28.4 45.2 12.9
Finland 32.9 70.8 23.3
Portugal 22.7 35.2 8.0
Ireland 44.0 61.3 27.0
Average Unweighted 31.9 50.5 16.1

Weighted by GDP 25.5 52.8 13.3
Standard Unweighted 14.1 11.0 7.3

deviation Weighted by GDP 10.8 25.2 5.9
Source: IMF Direction of Trade Statistics (1997 data).
Note: GDP weights are obtained from national GDP converted into 1997 ECU exchange rates.
To summarise, these arguments suggest that the introduction of the Euro is

likely to bring about a convergence of the monetary policy transmission
among participating countries, if attention is focused on exchange and
interest rate effects.
2 The lending channel: how will it change in Stage Three?
According to Kashyap and Stein (1994), three conditions must be present for
such channel to operate. First, monetary policy shocks must affect the overall
size of the banks' balance sheets ± i.e. money demand must be interest elastic;
second, loan supply must respond to money demand changes; third, loans
must be `special' ± that is, not easily substitutable by different forms of
®nancing for the private sector. On all these aspects, signi®cant changes are in
process in the Euro Area.
In recent years, ®nancial innovation has been affecting the nature and
properties of money demand in several countries. New ®nancial instruments,
close substitutes to money, have enabled banks to lower the burden of reserve
requirements and to increase the ¯exibility and diversi®cation of their balance
sheet's liability side. Often, new bank debt instruments have been introduced,
or existing ones have been modi®ed, in response to tax or regulatory changes.
Discussion 205
The growing importance of near-money instruments such as certi®cates of

deposit, repurchase agreements, short-term bank bonds, etc. has induced
central banks to broaden the de®nition of their monetary aggregates. The
M3 de®nition adopted by the ECB includes all bank liabilities with a
maturity up to two years; within this maturity span, banks may be able to
compensate a policy induced shortage of sight deposits by raising funds in
the near-money segments, thereby reducing the central bank's ability to
control the bank balance sheets and the supply of loans.4 This may actually
not be true for all individual banks or bank categories, as an active liability
management may be possible only for the more sophisticated and larger
players. However, the process of bank concentration experienced in recent
years in Europe (Table 5.2, p. 177) will tend to spread this phenomenon
across the area, though perhaps at different paces in each national market.
A parallel process of liability diversi®cation is likely to take place, over time,
also for non-®nancial ®rms. Though only very preliminary signs are yet
apparent, developed and diversi®ed private securities markets are likely to
emerge as a consequence of EMU, as investors will look for new instruments of
risk taking and diversi®cation. This should go in the direction of reducing the
dependence of corporate borrowers on bank loans as a form of ®nancing.
Finally, one should not underestimate the contribution that EMU will
provide in enhancing cross-border competition in bank lending. With the
elimination of currency risk, borrowers will increasingly be able to search for
the best credit conditions available to them across the whole area. In practice,
informational asymmetries and other tax and regulatory obstacles will
continue to be important factors in cross-border lending, speci®cally in
relation to small lenders and borrowers. However, competition will progres-
sively erode these segmentations and tend to give rise to broadly uniform
credit conditions for homogeneous classes of borrowers,5 thereby mitigating
any differences that exist, at the national level, in the impact of bank lending
behaviour in the transmission of monetary policy.
3 Two puzzling empirical results
Before concluding, I turn to the empirical evidence presented in the chapter.

I do it brie¯y because such evidence is admittedly suggestive, and therefore
should be taken more as an indication of how and in what direction further
research should proceed. Nevertheless, two speci®c results should, in my
view, be singled out as particularly surprising and worthy of deeper
investigation.
First, looking at the indexes presented in Table 5.5 (p. 181), and sumarised in
Table 5.8 (p. 187), countries belonging to the English-type legal family ±
notably, the United Kingdom ± emerge as being characterised by a
comparatively low effectiveness of monetary policy. This is in sharp contrast
with a commonly held view that Anglo±Saxon countries' household sectors
should be relatively more sensitive to interest rate changes. At least two factors
justify this presumption: the higher relevance of consumer credit and housing
mortgages, that tend to strengthen liquidity and income effects following
short interest rate adjustments, and the higher incidence of shares in private
portfolios, that give rise to stronger wealth effects. The empirical literature is in
agreement with this presumption: for example, the detailed cross-country
study by the BIS (1995) provides convincing empirical backing for it, based on
structural econometric models. Again, this puzzle may partly be explained by
the fact that this chapter concentrates solely on bank lending effects,
disregarding the interest and exchange rate channels.
Second, I found somewhat confusing that the signs of the predicted
in¯uence of the legal family on the effectiveness of monetary policy differ in
sign, in Table 5.8, according to whether the impact on output or in¯ation is
considered. Since all indicators analysed in the chapter refer to effectiveness of
monetary policy in affecting aggregate demand, one would expect the effects
on prices and quantities to have the same sign, and to vary in relative size
according to the nature of the in¯ation±output tradeoff.
4 Conclusions
As I stated at the outset, this is an interesting chapter, fun to read, one that
stimulates thinking. Its central statement, that the legal system may be an
important factor shaping the transmission of monetary policy, is convincing
and justi®es further research.
However, its empirical results and policy conclusions regarding the Euro
Area should be looked at with considerable caution. Relevant areas of the
overall monetary transmission process are not given explicit consideration, in
order to focus speci®cally on bank lending behaviour and on the nature of the
borrower ± lender relationships. Though this simpli®cation may be exposi-
tionally convenient, it should be borne in mind that it could change not only
the size, but also the sign, of the results. Some puzzling empirical estimates
mentioned above may derive precisely from this somewhat unbalanced view
of the transmission process.
The evolution in process in the European banking markets does, in my view,
deserve more attention than it receives here. Restructuring, consolidation and
competition are likely to change considerably the European banking land-
scape and the way bank behaviour affects the transmission process, both in
the aggregate and across different countries, sectors and classes of borrowers.
The `credit channel' in the Euro Area may well have to be studied anew,
building on existing knowledge but taking into account the relevance of the
regime change we are living through. Indeed, for the Eurosystem this may be
one of the most challenging and policy-relevant research issues in the years to
come.
Discussion 207
Notes
1. The views expressed here are the author's, and do not necessarily re¯ect those of the
European Central Bank. I am grateful to Benoit Mojon for useful discussions.
2. The modern literature on the `lending view' has elements in common with the old
one on the àvailability doctrine' (see Roosa, 1966). The main difference is that the
new view does not necessarily rely on credit rationing: loan market effects can take
place in equilibrium, through changes in the spread between the loan rate and the
money market rate.
3. I am assuming here that households do not fully internalise future tax liabilities, ±
or, equivalently, that the public debt is at least partly perceived as net wealth by
consumers.
4. According to econometric estimates (Coenen and Vega, 1999) the area-wide M3
aggregate is not controllable in the long run via interest rate changes only.
Controllability is restored if one also considers the effects on output and prices.
5. Since the mid-1990s, a sharp convergence was indeed observed in lending rates
across EMU countries. Such convergence was, however, related not only ± or not
much ± to cross-border competition in lending, but to a large extent to the general
convergence of in¯ation and interest rates.
References
Angeloni, I. and Rovelli, R. (1998) Monetary Policy and Interest Rates, London, Macmillan.
Bank for International Settlements (BIS) (1995) Financial Structure and the Monetary Policy
Transmission Mechanism, Basle, BIS, March.
Monetary Transmission' American Economic Review, 82, 901±21.
Coenen, G. and J. L. Vega (1999) `The Demand for M3 in the Euro-Area', European
Central Bank, Working Paper Series 6, September.
Gertler, M. and S. Gilchrist (1993) `The Role of Credit Market Imperfections in the
Monetary Transmission Mechanism: Arguments and Evidence', Scandinavian Journal
of Economics, 95, 43±64.
Hubbard, R. G. (1995) Ìs There a ``Credit Channel'' of Monetary Policy?', Federal Reserve
Bank of St Louis Economic Review, 77, May±June, 63±77.
Kashyap A. K. and J. C. Stein (1994) `Monetary Policy and Bank Lending', In N. G.
Mankiw (ed.), Monetary Policy, Chicago, University of Chicago Press for NBER.
ÐÐÐÐ (1997) `The Role of Banks in Monetary Policy: A Survey with Implications for the
European Monetary Union', Federal Reserve Bank of Chicago Economic Perspectives,
September±October, 2±18.
La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. W. Vishny (1997) `Legal
Determinants of External Finance', Journal of Finance, 52, 1131±50.
ÐÐÐÐ (1998) `Law and Finance', Journal of Political Economy, 106, 1113±55.
Rajan, R. G. and L. Zingales (1998) `Which Capitalism? Lessons from the East Asian
Crisis', Bank of America, Journal of Applied Corporate Finance, 11, 40±8.
Roosa, R. V. (1966) Interest Rates and the Central Bank; reprinted in Richard S. Thorn,
(ed.), Monetary Theory and Policy, New York, Random House, 559±83.
6
Differences Between Financial Systems

in European Countries: Consequences
for EMU1
Reinhard H. Schmidt
1 Introduction
There are three main issues which must be discussed if one wishes to answer
the question which this chapter addresses:
(1) Which features of a ®nancial system are important for monetary policy
or, in other words, how is monetary policy conducted, and how does it
affect the real economy, and how and to what extent does this depend on
the speci®c features of an economy?
(2) How different are the ®nancial sectors ± or more generally the ®nancial
systems ± in Europe?
(3) If signi®cant differences existed between countries, would this have
consequences for how monetary policy should, and can, be conducted in
a common currency framework?
I want to warn readers at the outset that I am not a ®nancial

macroeconomist. Because this is not my ®eld of specialisation, my knowledge
of how monetary policy is conducted and how it affects the real economy ±
that is, of the so-called transmission mechanism ± is only super®cial. I can only
hope that this does not invalidate all that I would like to offer in my attempt to
ful®l the role assigned to me as a microeconomically oriented scholar in
banking and business ®nance. In order to clarify the basis on which I stand, I
want to ®rst present two simple ideas about why differences between the
national ®nancial systems in Europe might matter for monetary policy in
Èuroland'.
In essence, I see monetary policy and its effects as follows. There is a central
bank which has the ability to provide a monetary policy impulse: the central
bank can, for instance, raise the short-term interest rate or reduce the quantity
of central bank money in the economy (and will typically do both at the same
time). This impulse has immediate consequences for the interest rate structure
and the quantity of money (according to some speci®c de®nition of a
208
Differences Between Financial Systems in European Countries 209
monetary aggregate) in the economy and for the banking sector. To the extent
that banks re®nance themselves, or hold voluntary or compulsory reserves, at
the central bank, they have fewer funds to offer to their potential borrowers
and/or lending becomes more costly to them. The resulting change in the
situation of the banking sector has, in turn, consequences for the real
economy to the extent that the ®rms, households and government bodies that
want to borrow from the banks have less credit available, or have to pay more
for the funds borrowed from the banking sector. The restrictive monetary
policy impulse thus works itself through the system and determines aggregate
demand and ultimately output and income, and possibly the price level as
well. Some details of this process have to be left for further discussion, but
evidently I assume that monetary policy is effective, at least in the short run.
The second idea is taken from a paper by Dornbusch, Favero and Giavazzi
(1998). Differences between the national ®nancial systems might matter
because different national economies might react differently to monetary
policy impulses of a given kind and size. If this is so, it might be so because of
differences in the ways in which central bank action in¯uences the respective
®nancial sectors and in particular the banking sectors, or because of differences
in the ways in which a change in the situation of the ®nancial sectors affects
the real economy. If there are two countries, A and B, in a monetary union,
with A being more sensitive to monetary policy impulses than B, then a given
monetary policy measure might be too strong for country A and too weak for
country B even if it is just right for the intended effect on the (weighted)
average of the two economies in terms of, say, their in¯ation rates.
If I may claim any comparative advantage at all for discussing the
consequences of national differences for EMU, it has to do with the second
question ± that is, the differences between the ®nancial systems in Europe.2
However, here I must also make two reservations:
From the way I have formulated the title of my chapter and the questions
listed above, it can be seen that I will not con®ne myself to looking at
differences between ®nancial sectors, but wish to ± and will later show that I
need to ± discuss differences between national ®nancial systems. This is not
simply a terminological difference, but a difference in substance. I use the term
`®nancial sector' in a narrow way: it denotes those specialised institutions such
as banks, pension funds, securities markets, etc. which provide ®nancial
services to the non-®nancial sectors of the economy, and to the ways in which
these institutions are shaped and managed and how they operate and are
regulated. The term `®nancial system' encompasses not only the ®nancial
sector, but also the real sectors to the extent that they demand the ®nancial
services of the ®nancial sector and also to the extent to which they forgo using
the ®nancial sector, as well as the interaction between the demand for and the
supply of the services of the ®nancial sector. Thus, for instance, the extent to
which internal ®nancing of investment takes place, the extent to which saving
takes the form of real investment, the extent to which banking services are
210 Reinhard H. Schmidt
appropriate to the demand for them, etc. are features of a given ®nancial
system.3
The second reservation refers to the countries which I will take into
consideration. Anticipating a likely future course of events, I will assume that
the United Kingdom is already part of the European Monetary Union (EMU).
Moreover, my empirical references will be mainly restricted to the three largest
®nancial systems in the ènlarged EMU' ± namely those of France, Germany
and the United Kingdom. Even if the Italian economy were larger than that of
Great Britain, I would wish to include that of the United Kingdom because, as I
will describe below, its ®nancial system differs markedly from those of the
continental European economies, and from that of Germany in particular. An
understanding of these differences helps to underscore the heterogeneity that
makes for national diversity, and thus for possible problems of a common
monetary policy, in the EMU.
The chapter is structured as follows: In Section 2, I shall discuss the
transmission mechanisms. Given my background, it appears almost self-
evident to me that the so-called `credit channel' is important,4 though not so
much as an alternative to the `conventional' channels of the transmission of
monetary policy, but rather as a complement to them. The discussion of the
various channels serves the purpose of providing the criterion for selecting
those aspects of national ®nancial systems which I will then go on to
characterise in some detail. In Section 3, I will argue that the ®nancial systems
± as well as the ®nancial sectors ± of the three countries are vastly different, and
point out the main differences. In Section 4, I will bring the two lines of
reasoning together and argue that, in spite of what I perceive as important
differences, I do not think that these differences matter very much for the
design and conduct of a common monetary policy. The main reason for this is
that what I consider to be the characteristic features of the different ®nancial
systems ± though not of the different ®nancial sectors ± might have two effects
on the functioning of monetary policy which tend to offset each other.
2 Transmission mechanisms or channels of monetary policy
The classi®cation of channels

As the name suggests, a transmission mechanism is a conceptual or formal
model of the ways through which monetary policy in¯uences the real
economy. These ways are complicated and, as far as I can judge, still
imperfectly understood. In the relevant literature,5 four different transmission
mechanisms or channels are distinguished:
(a) the interest rate channel

(b) the channel of relative prices
(c) the exchange rate channel, and
(d) the credit channel.
The interest rate channel (a) and to a certain extent also the channel of
relative prices (b) are standard elements of what may be regarded as the
traditional view of the transmission mechanism. The exchange rate channel
(c) is not relevant in the context of the present chapter, as the ± enlarged ±
EMU area is a very large economy for which exchange rates are not a terribly
important factor because external trade accounts for only a relatively small
share of total GDP. The credit channel (d) is the `newcomer' in the market of
ideas and by now the main `competitor' to the incumbents (a) and (b).
The interest rate channel and the channel of relative prices

The interest rate channel is based on the conventional Keynesian IS ± LM
model. According to this model, the central bank determines short-term
interest rates. With given and unchanged expectations about the in¯ation rate,
this also has an effect on real longer-term interest rates, which determine the
investment decisions of pro®t maximising ®rms, as the ®rms compare
marginal internal rates of return on their investment projects with `the cost
of capital' when they decide whether to invest or not. Similar considerations
apply to certain consumption decisions,6 so that an unexpected change in
that interest rate which the central bank can set in¯uences aggregate demand
and ultimately also output. Given that prices are in¯exible only in the short
run ± not only by de®nition, but also in reality ± an expansionary or restrictive
monetary policy impulse provided by the central bank loses its effect on the
real economy over the course of time when prices start to react.
In order for the interest rate channel to function, two conditions must be
met: for one thing, monetary policy must not only affect short-term interest
rates but also (real) medium to long-term rates, and, for another, investment
decisions must be interest-elastic. Differences between ®nancial sectors and,
more generally, ®nancial systems can translate into differences in terms of the
strength of the links between nominal short-term and real longer-term interest
rates and in terms of the strength of the links between these rates and
investment and consumption decisions. The effects of monetary policy are
stronger if ®nancial contracts and, in particular, the terms over which interest
rates are contractually ®xed are shorter and thus more easily adjusted. Another
factor which may in¯uence the effectiveness of monetary policy, as its
working is described in the interest rate channel view, is the extent to which
central bank money is being used in the economy.
The channel of relative prices (b) ± which is also sometimes called the àsset
price channel', which is presented in two different versions in the literature,
namely in the `monetarist' version and that developed by Tobin (1969) ±
assumes that central banks in¯uence the composition and/or prices of the
assets which are held in the portfolios of economic agents. An unexpected
monetary policy impulse disturbs the equilibrium composition of the
portfolios and induces attempts to adjust their composition. Ultimately, an
expansion of the monetary base by open market operations (OMOs) leads to

more demand for securities, to rising security prices and thus falling interest
rates and possibly also to an increase in consumption and investment
expenditure. The extent to which these effects on asset prices merely lead to
price-level changes, or also to real effects, depends on the rigidity of prices and
the `disturbances' which may arise from changes in the expectations
concerning the future monetary policy. The two versions mentioned above
differ with respect to the asset categories which the agents try to bring into
balance, and their respective rates of return and also as regards the extent to
which the process of the restructuring portfolios directly affects the demand
for real versus ®nancial assets and thus investment decisions.
According to the proponents of the asset price channel, the main instrument
of the central bank is its in¯uence on the quantity of central bank money in the
economy. This suggests that different procedures in which central banks
operate in practice, and, in particular, differences in terms of the types of
®nancial assets which are eligible as reserves, may lead to differences in the
effectiveness of monetary policy. However, even also including the United
Kingdom, the differences between the various economies in Europe which
existed in these respects in the past, are giving way to a common approach.7
Another aspect, and one in which the various European economies still differ
considerably and which is important for this channel, is the extent to which
the agents hold ®nancial assets whose prices may vary in reaction to central
bank policy. The larger the share of such ®nancial assets in agents' portfolios,
the greater will be the likelihood that central bank impulses will work in the
ways postulated by the advocates of the asset price channel.8
It appears plausible that monetary policy is able to have an effect on the real
economy primarily because of interest effects and also, to a certain extent,
because of asset price effects. However, the recent empirical literature argues
that in reality the effects of monetary policy, in particular those of a monetary
contraction, are stronger and of a different pattern than those which could be
expected if the interest rate and asset price channels were the only relevant
mechanisms, and also that they exhibit a different pattern than the effects one
would expect to encounter if these were the only pertinent mechanisms.9 At a
theoretical level, the interest and the asset price channels assume that the
agents in the economy behave in an overly mechanistic fashion, and fail to
address directly the question of how the ®nancial sector, and, in particular, the
banks, react to monetary policy impulses. These two weaknesses suggest that
looking only at the traditional channels might prevent one from acquiring a
deeper understanding of the implications of intercountry differences for a
common monetary policy.
The credit channel

At the heart of the so-called credit channel are those aspects which are largely
left out by the two traditional channels. Because of the information and
incentive problems which are widely discussed in the current corporate

®nance literature, ®nancial systems do not function in a frictionless manner,
and for many non-®nancial ®rms external ®nancing is simply dif®cult to
obtain and more costly than internal ®nancing. The cost difference is called
the èxternal ®nance premium'. In the relevant theoretical and empirical
literature this external ®nance premium is assumed to be not only an expense
incurred in addition to the basic interest costs, but also a positive function of
the interest rate. In other words, if the central bank raises or lowers `the interest
rate', the external ®nance premium will also go up or down.10
The second aspect which is highlighted in the credit channel view of the
transmission mechanism comprises the availability of bank credit and the
speci®c quality ± i.e. the limited substitutability ± of bank credit. Working in
combination, the external ®nance premium and the availability of bank credit
strengthen the effectiveness of monetary policy considerably, and it seems
thoroughly plausible to assume that these two factors may differ much more
between countries than those at which the traditional views of the
transmission mechanism focus.
As this brief introduction suggests, the credit channel can be broken down
into two separate but complementary channels. The ®rst one, the broader of
the two, is called the `balance sheet channel'. It focuses on the ability to
borrow. The external ®nance premium for a given borrower is determined by
its ®nancial position, in particular by its net worth, and the value of the
collateral which it can provide. The borrower's ®nancial position is
in¯uenced by the monetary policy and by the business cycle. A restrictive
monetary policy raises the interest rates, reduces the cash ¯ow of ®rms and
depresses net asset values of borrowers and the value of their collateral, and
thus may severely restrict their ®nancing options and raise the premium. The
likely consequence is a reduction of investment and in particular of
investment in working capital, which is typically ®nanced by short-term
bank credit. Owing to economy-wide accelerator effects, the impact of a
monetary contraction may not only be stronger, it may also last longer than
the traditional view suggests.11
The second `branch' of the credit channel is the more narrowly de®ned
bank lending channel. Its proponents proceed from the highly plausible
assumption that the central bank is in a position to limit the quantity of credit
which the banking sector can provide to borrowers. A restriction or rationing
of bank credit in turn restricts the scope of ®rm investment; and this is all the
more likely the less bank credit can be substituted by other sources of funding
at the bank and ®rm level. In claiming that bank credit is indeed dif®cult and
in some cases even impossible to substitute as it provides a certain liquidity
insurance, advocates of the credit channel of the transmission mechanism
borrow heavily from the recent advances in the theory of ®nancial
intermediation which shows why `bank loans are unique'.12 Banks are
specialists in lending to ®rms in those cases in which it is important to
monitor the borrower carefully ± or, in other words, in overcoming

information and incentive problems.
The balance sheet channel and the bank lending channel, moreover, interact
in such a way that the effects of monetary policy on the ability of ®rms to borrow
from banks and on the ability of banks to lend, reinforce each other. This makes
their relevance for the effects of monetary policy all the greater. As these brief
explanations suggest, it would be wrong to consider the credit channel as an
outright alternative to the interest rate channel. Instead, the effects of borrowing
capacity and the availability of bank credit which this channel emphasises
reinforce those effects which have traditionally been assumed to exist and to
underlie the transmission of monetary policy into the real economy. But both
for practical monetary policy and for the problem which is discussed in this
chapter ± namely that of possible consequences of differences between national
®nancial systems ± it is important to know more than merely that monetary
policy has an effect. One needs to know why it matters and how it affects the real
economy in order to be able to determine the direction and the strength of
policy measures.13
The credit channel view suggests a list of items which can be used to check
which elements of a ®nancial system might be particularly relevant when one
tries to analyse the implications of differences in ®nancial systems between
countries for monetary policy. This list would be too comprehensive to
develop here, and I will thus restrict myself to indicating classes of factors.
The central bank and the money market structure are the ®rst category.
Some of the key questions here are: How is monetary policy implemented?
Who interacts directly with the central bank? How deep and liquid are the
markets for short-term securities like CDs, commercial paper and government
bills? What scope for administrative interference in the short-term operations
of banks do the government and the central bank have?
At the level of the banking sector, some of the relevant issues are: What are
the most prevalent types of banks? What is the typical relationship between
banks and ®rms? What are the legal forms and ownership patterns of banks?
What is the level of concentration and competition in the banking sector?
What is the nature of the relationship between banks and non-bank ®nancial
intermediaries (NBFIs)? Are the NBFIs competitors or subsidiaries of banks?
What is the asset and liability structure of the banks' balance sheets?
NBFIs and capital markets constitute the main alternatives to banks. They
need to be looked at if we wish to determine the extent to which ®rms and
banks are able to circumvent the effects of a restrictive monetary policy which
would be transmitted through the bank lending channel. As far as NBFIs are
concerned, some of the most important factors are (a) the nature of their
relationship to the central bank and to the banks, (b) how important they are
in the ®nancial sector, and (c) their asset and liability structures. As far as
capital markets are concerned it is important to know which markets exist,
who has access to them and how deep and how liquid they are.
Finally it would be important to look in detail at the sector of the non-

®nancial ®rms and at the household sector to assess how the actors in these
sectors are affected by monetary policy measures of the central bank. Key
questions here would be: How are they typically ®nanced? What is the
prevailing mode of holding liquidity? How many of them have access to
security markets or to foreign sources of funding? And, last but not least, how
are they governed and what is the time horizon of their strategies?
If the goal is to conduct a reasonably comprehensive comparison of
®nancial systems to assess the consequences of national differences for a
common monetary policy, then the view one adopts concerning the
transmission mechanism of monetary policy evidently has a strong in¯uence
on the speci®c choice of aspects which must be examined. The credit channel
implies a much longer list of factors than the traditional channels. The
characteristics of some of these items in the ®nancial systems of France,
Germany and United Kingdom will be described on p. 218 and in the
Appendix. However, even more important than the individual items in this
list ± which is not intended to be complete ± is the question how they are
related to each other. This is a topic to which I will return in the next section.
In my opinion, the credit channel view is extremely helpful for an analysis of
the consequences which differences in entire ®nancial systems have for the
topic under discussion here ± an assertion whose validity I will seek to
demonstrate in Section 4.
3 How different are ®nancial systems in Europe?
How different can ®nancial systems be?

To start this section, I want to recall the de®nition of the ®nancial system
provided in the Introduction. The ®nancial system includes the ®nancial
sector as the provider of ®nancial services as well as the real sectors of the
economy in so far as they demand ± or, as the case may be, fail to demand ±
these services, and the complex relationships between the ®nancial and the
non-®nancial sectors. Evidently, it does not only depend on the ®nancial
sector but also on the real sectors and the relationships between them how
monetary policy affects the real economy. Thus, the broad de®nition is
appropriate for any analysis of monetary policy.
This de®nition is also important for another reason. It helps to get a more
comprehensive picture of the extent to which the ®nancial systems of
countries differ, and the respects in which they differ, than could be obtained
using the narrow concept of the ®nancial sector. Indeed, as I will argue in this
section, both ®nancial sectors and ®nancial systems differ between countries,
but the latter are more dissimilar than the former, and differences in other
parts of the ®nancial systems ± i.e. in parts other than the ®nancial sectors ±
could be highly relevant for the design and conduct of a common monetary
policy. Thus the broad de®nition of the ®nancial system is particularly
Financial System
(1) (2)
Financial Financial
sector patterns
(Role of (Saving and
institutions and financing)
market)
(3) (4)
Corporate Corporate
governance strategy
(Roles of various (Ability regarding
stakeholders) fundamental
change)
Figure 6.1 The ®nancial system
appropriate for studying the problems of a common monetary policy in

different countries, as it points to the fact that those aspects of the real
economy which are important for the working of the transmission mechan-
ism and differ between countries may correspond in a systematic or non-
accidental way to the differences between the ®nancial sectors. This
correspondence is crucial for the effects of a common monetary policy.
If I make this claim, I am obliged to take a closer ± which is at the same time
also a more abstract ± look at the concept of a ®nancial system.14 The broadly
de®ned ®nancial system is an open system; there is no clear factual criterion to
determine what belongs to it and what does not. Rather, the boundaries of the
system at large are drawn by the observer who, in a given case, seeks to address
a certain problem in the most fruitful way. For the purpose of this chapter, one
can distinguish four closely interrelated sub-systems which are, in turn,
composed of several elements. Figure 6.1 shows these sub-systems, which are
at the same time elements of the larger system, the ®nancial system. They
include, ®rst of all, the ®nancial sector. The second sub-system consists of the
surplus and the de®cit units in so far as they provide funds to the ®nancial
sector or obtain funds from it, and thus of the savings behaviour of households
and of the ®nancing of business or corporations and governmental bodies,
and of the ®nancial instruments used by the parties to ®nancial contracts. It is
advisable to include corporate governance as the third sub-system and
corporate strategies and structures ± for short: the business system ± as the
fourth. Note that there is no need to de®ne these sub-systems in such a way
that they do not overlap.
It is imperative to see how these sub-systems are composed and function
individually, and how they are related to each other. As the colloquial use of
the term `system' suggests, a system is more than a collection of elements. For a
system to be `really a system', its elements must `®t together'. In economic life,
one can assume that it has positive consequences if this is the case, and
negative consequences if the elements do not ®t. One can speak of ìmportant'
differences between ®nancial systems, if not only the main elements of which
they are composed, but also the way in which they are related, are different.
The interesting thing about the sub-systems and their interrelationship,
and thus the interesting thing about the entire ®nancial system, is that they
are composed of complementary elements. Elements of a system are called
`complementary' (to each other) if they mutually increase the `bene®t' they
yield in terms of whatever the objective function or the standard for
evaluating the system may be, and mutually reduce their disadvantages or
`costs'.15 A system is called consistent if its complementary elements indeed
take on the values which make the system attain a local optimum. Systems
of complementary elements typically have more than one optimum, and
the local optima are clearly distinct con®gurations of the values of the
elements.
Financial systems are systems in this speci®c sense. The complementarity of
their elements and the economic bene®ts which a consistent ®nancial system
can be assumed to produce are the factors which account for the tendency of
countries' ®nancial sectors and non-®nancial sectors to co-vary in a systematic
way. To illustrate the concepts of complementarity and consistency in the case
of ®nancial systems and also to lay the foundation for the next argument, I
would now like to take a look at common two-way classi®cations from the
literature for the four sub-systems shown in Figure 6.1.
(1) For the sub-system `®nancial sector' it is common to distinguish between

a bank-dominated and a capital market-dominated variant based on
certain measures of the size and ìmportance' of banks and capital
markets, respectively.
(2) The second sub-system refers to how savers hold their ®nancial wealth
and the sources from which corporations obtain the bulk of the funding
for their investments. This suggests a similar two-way classi®cation
distinguishing a bank- and a capital market-oriented system.16
(3) With respect to the third sub-system, corporate governance, there are
two-way classi®cations according to different criteria depending on
whether the safeguarding of the interests of owners, of creditors, or of the
staff is involved. The details need not interest us here.17 Suf®ce it to say
that these classi®cations can be brought together under the heading of
ìnsider-' and òutsider-controlled' corporate governance systems.18
(4) Lastly, ®rms can be classi®ed according to their business systems into two
groups: those which make ample use of ®rm-speci®c human capital, rely
to a large extent on long-term and trust-based relationships ± or implicit
contracts ± with their various `stakeholders' and partners, and are more
suited to undertake strategic adjustments through longer sequences of
small changes, and those with the opposite characteristics.19
Looked at individually, each of these eight speci®c sub-systems can be

interpreted as consistent systems of complementary elements. Owing to space
limitations, I cannot demonstrate this here. Instead, I will concentrate on the
question of how the four sets of sub-systems are related to each other: The sub-
systems are themselves complementary elements of the overall system; and
only two speci®c combinations constitute consistent systems. As far as the ®rst two
elements are concerned, this seems plausible. Even though there may be an
empirical problem in demonstrating this correspondence, which I shall
discuss below, one would tend to think that there is a one-to-one
correspondence between the two classi®cations in these two sub-systems: a
system in which savers entrust a large fraction of their wealth to banks or
closely bank-related non-bank ®nancial intermediaries should also be one in
which the banks are the dominant element in the ®nancial sector; and capital
market dominance should equally be re¯ected in the patterns of saving and
®nancing and the size and role of the capital market as an institution.
Moreover, a bank-dominated ®nancial sector and a bank-oriented system of
savings and ®nancing also `®t', or correspond in functional terms, to an
insider-controlled governance system and a business system with much ®rm-
speci®c human capital, many implicit contracts and a tendency towards
gradual change.20 The set of other `values' for the four sub-systems as elements
also constitutes a consistent ®nancial system. It is characterised by a greater
importance for capital markets than for banks in the ®nancial sector, more
saving and ®nancing via the capital market than with and from banks,
outsider-controlled corporate governance and highly ¯exible corporations
which are able to undertake strategic adjustments with `big leaps'. Evidently,
these two systems are clearly distinguishable types of ®nancial systems or, to
use Williamson's suggestive term, `discrete alternatives'.21
So at least in theory, ®nancial systems can be fundamentally different. And
since the four sub-systems in¯uence each other in their functional values,
entire ®nancial systems can be `more different' ± i.e. different in a more
fundamental way ± than ®nancial sectors alone might.
How different are those in Europe really?

The fact that one can distinguish two types of ± consistent ± ®nancial systems
at a conceptual level does not imply that one could also ®nd these different
types in reality, nor does it imply that real ®nancial systems do in fact
correspond to this typology. Therefore, the questions of whether real ®nancial
systems have complementary elements, and thus whether they are `really
systems' ± and, if they are, whether they are also consistent systems ± are
empirical questions. It should also be noted that the theoretical distinction
does not tell us whether the ®nancial systems of the major European
economies are different. Thus, we shall now turn to the empirical issues. It
goes without saying that ± not only owing to the limited space available ± only
a very brief discussion of this complicated subject can be provided here;
however, I hope that by drawing attention to a few key aspects I will be able to
make my point. It is equally clear that one cannot expect to ®nd a perfect
correspondence between a given type of ®nancial system and a given real
®nancial system ± i.e. that of a speci®c country.
Nevertheless, the classi®cation of ®nancial systems suggests differences in
reality.22 The example which immediately come to mind are the German and
the British ®nancial systems ± and this is the main reason why I anticipate a
likely future course of events and assume that the United Kingdom is already
part of Èuroland'.
The German ®nancial system is of the ®rst type. The German banking sector
is big and some banks are important and powerful, whereas capital markets
and capital market-oriented institutions such as pension funds are ùnder-
developed', as can be measured by indicators such as the total assets of banks
and stock market capitalisation as percentages of GDP, respectively.23
Corporate governance in Germany functions mainly through internal
mechanisms and involves ìnsiders' to the corporations ± or, in other words,
people and institutions which have long-term interests in safeguarding their
speci®c relationships with a corporation and that are typically better informed
about its prospects and problems than anonymous market participants could
be.24 There does not seem to be an active market for takeovers. The role of
banks, of other corporations and of employees having codetermination rights
in the governance of corporations epitomise this system, and despite the
widely publicised declarations to the contrary, the maximisation of share-
holder value is not the dominant objective of most large German ®rms. Even
though this may be more speculative, one can also add the observation that, at
least on an economy-wide level, German ®rms are more woven into nets of
implicit contracts, that labour turnover is lower and that the corporations are
less able and willing to undertake abrupt adjustments to changing circum-
stances than British corporations. Thus it is fair to say that at least as regards
the ®rst, the third and, to a certain degree, the fourth element (or sub-system),
the German ®nancial system is consistent.
Without doing too much violence to a reality which is of course much more
complex, one can characterise the British ®nancial system as one in which
core elements ± or sub-systems ± (1) and (3), and possibly also (4),25 take on
values which are precisely the opposite of those determined for Germany. In
the United Kingdom, the relative importance of banks in the ®nancial sector is
not as great; capital markets and institutions ± notably pension funds, which
are capital market-oriented and in most cases independent of the banks ± play
an important role; there is no formal and no de facto codetermination;
and ®nancial intermediaries are scarcely involved in the governance of
corporations.26 The corporate objective is unrestricted maximisation of pro®ts
or shareholder value, and given the nature of the UK ®nancial system there is
also no reason why this should be otherwise. The basic mechanism of
corporate governance is the takeover market, and the entire corporate
governance system is clearly an outsider control system.27 Overall, the British

®nancial system relies very much on market mechanisms and less on
institutions in the conventional meaning of the term ± i.e. it relies less on
groups of people who interact directly with each other over a long period and
not only have a certain degree of mutual commitment but also ®nd that their
interaction gives rise to intense con¯icts over important issues which are,
however, resolved internally and ultimately in a peaceful way.28
Despite far-reaching changes in the regulatory and competitive environ-
ment in which ®nancial sectors and non-®nancial corporations in Germany
and the United Kingdom operate, it is fair to say that the characteristic features
of these two ®nancial systems, as far as I have characterised them up to now,
are surprisingly stable. Many details of the ®nancial systems have been
modi®ed, but their fundamental or structural features have not changed over
the last twenty years ± and, indeed, they have probably remained essentially
unaltered for much longer than that.
In the case of the French ®nancial system, it is more dif®cult to ®nd a
unifying `logique'. At least the efforts in this area which I have undertaken with
colleagues and students ± for some time now, we have been trying to locate the
French ®nancial system along the spectrum of market and institution-based
(or Anglo±Saxon and German-type) systems ± have not been successful.29 This
is so for at least two reasons. One is the strong role which the French state (or
government) has played in the ®nancial system over the years. It is extremely
dif®cult to integrate the extent and the forms of state or government
intervention into the two-way classi®cation which was described on p. 218, or
to invent a different classi®cation incorporating the additional dimension of
state activity.30 The second reason is that the French ®nancial system and its
four sub-systems have experienced dramatic and fundamental changes during
the last two decades. If one looks at individual elements, it appears that these
changes have already converted the French ®nancial system from being of the
German type ± except for the pervasive state intervention ± to one of the
Anglo±Saxon type.31 This may indeed be the case. But in the past the French
®nancial system had a number of elements which were certainly not
congruent with the German model or the German ®nancial sector type, and
today it includes several elements which do not `®t' the general structure of the
Anglo±Saxon model or ®nancial sector type.
Referring not only to European ®nancial systems, but also to those of Japan
and the United States, Figure 6.2 summarises the characteristic features of
consistent ®nancial systems. The tables in the Appendix provide empirical
support for my brief summary description of the three ®nancial systems.
Nevertheless, it would be good to have more empirical evidence to
underscore the profound differences between Germany and the United
Kingdom and to support my view that the system in France does not conform
to either of the two `polar' models and is at present unstable in so far as it is
undergoing a process of rapid change. In particular, additional evidence is
Stability of business environment
BE Higher Lower BE
Germany
FS FP Japan
– Households hold
– Banks play primarily bank deposits
– Bank loans as main
Lower
Lower
dominant role
financial source of firms
Limited
coherence
CG CS
System-wide role of market mechanisms
– Control – Firms-specific
– Relationships investments
– Inside influence – Continuous
– Insider control adjustments
USA
FS FP
UK
– Organised – HHs hold claims
capital markets on funds and securities
play dominant – Market-based
role firm financing
Higher
Higher
Limited
coherence
CG CS
– Liquidity – Marketable
– Arm's length investment
– Outside option – Fundamental
– Outsider strategic shifts
control
BE Higher Lower BE
Figure 6.2 Consistent ®nancial systems

Note: FS = Financial sector; FP = Financial patterns; CG = Corporate governance
CS = Corporate strategy
called for in one respect: in my account of the ®nancial systems, I have

intentionally left out the second sub-system of ®nancial patterns. It is logical
to ask why, and of course the obvious answer is that it might not accord with
the general pictures of the systems. And while the savings part does not offer
any problem in this respect, the ®nancing part does. At least this appears to be
the case if one accepts at face value the results of an important strand in the
academic research about how ®rm investment is ®nanced. The ®ndings
presented in this literature are plainly inconsistent with the view underlying
my argument which is inspired by the theory-based assumption that there
should also be differences in the ®nancing patterns of the ®rms in the different
countries and that these should correspond to the differences between, both,
the respective ®nancial sectors and corporate governance systems. This
inconsistency has to be clari®ed if my claim that there are in fact fundamental
differences is to be regarded as valid.
In a series of highly in¯uential papers, Colin Mayer and others who followed
his line of research32 have arrived at the opposite conclusion. First of all, they
take the position that there are hardly any systematic differences between
national patterns of corporate ®nancing, and secondly that those differences
which do exist, are inconsistent with what one would expect on the basis of
theory and of the empirical features of sub-systems (1) and (3) in the various
countries. For instance, Mayer (1988, 1990) found in his early studies that
bank ®nancing is more important in the United Kingdom than in Germany
and that equity ®nancing is even negative in the United Kingdom. Corbett
and Jenkinson (1996, 1997) supported these results in more recent and much
more extensive studies. These ®ndings do indeed pose a puzzle. If they were
true, they would break the logical chain connecting the four sub-systems and
contradict the proposition that the German and the British ®nancial systems
are consistent. Moreover, they would invalidate the whole concept of ®nancial
systems as consistent sets of complementary elements.
From the perspective adopted in the present chapter, it is fortunate that the
methodology developed by Mayer and used by him and others has a ¯aw.
Mayer and his followers look at net ¯ows of funds between economic sectors
whereas they should have looked at gross ¯ows, because in the process of
aggregating over real and ®nancial investments and over time, netting
eliminates the relevant differences which might exist, and thus the role of
external debt ®nancing disappears almost completely. In his dissertation,
Andreas Hackethal (1999) has identi®ed this ¯aw, suggested a way to avoid it
and recalculated the ®nancing structures of ®rms in Germany, Japan and the
United States. As the ®rst columns per country (with the heading `gross') in
Table 6.1 show, his correction yields patterns of long-term ®nancing which are
consistent with expectations based on the structure of the entire ®nancial
systems. The second columns (`net') and the ®rst line concerning internal
Table 6.1 Financing patterns in various countries, 1970±96, per cent of physical
investment (gross: 1970±96; net: 1970±94)
Germany Japan USA UK

Sources of ®nance Gross Net Gross Net Gross Net Net
Internal funds 78.9 69.9 96.1 93.3

Bank ®nance 72.9 11.9 152.3 26.7 65.3 11.1 1 4.6
NBFI ®nance 9.3 n.a. 0.0 n.a. 24.6 n.a. n.a.
Bonds 7.0 ±1.0 14.3 4.0 48.5 15.4 4.2
Commercial paper 0.0 n.a. 5.7 n.a. 9.2 n.a. n.a.
New equity 3.8 0.1 3.8 3.5 14.9 ±7.6 ±4.6
Trade credit, others and n.a. 10.1 n.a. ±4.1 n.a. ±15.0 ±7.5
statistical adjustment
Note:
Net ®gures and those for Ìnternal funds' are taken from Corbett and Jenkinson (1997). They add up
to 100 per cent. These authors include both `NBFI ®nance' and `Commercial paper' in `Bank
®nance'(except for the United States where `Commercial paper' is included in Òthers').
Gross ®gures, which by construction do not add up to 100 per cent, are taken from Hackethal (1999).
Only long-term external funds are shown.
funds, for which the net and gross ®gures are identical, show the correspond-
ing net ®gures derived by Corbett and Jenkinson (1997) for comparison.33
The data needed to recalculate the ®gures for French and British ®rms, and
thus to provide a more accurate picture of their ®nancing patterns, are not yet
available, but it seems plausible that, given the similarity of the ®nancial
systems of the United States and the United Kingdom, those for the United
Kingdom will not differ too much from those for the United States. If this is
indeed the case, the claim that the British and the German ®nancial systems
are consistent and differ in a fundamental way can be upheld.
In another empirical study we have found further interesting evidence to
support the assumption of fundamental differences between the British and
the German ®nancing systems and the proposition that dramatic changes are
under way in France.34 In this paper, we have calculated various intersectoral
intermediation and securitisation ratios for the three countries over a span of
®fteen years. In contrast to the `conventional wisdom' that there is a general
tendency towards disintermediation and securitisation and that, overall, the
®nancial systems in Europe are becoming more similar, our study shows that
the levels of disintermediation and securitisation differ substantially between
countries and that, except for France, they are surprisingly stable over time. As
Figure 6.3 on p. 224 shows, the `liability±intermediation ratios of non-
®nancial companies', which measures the fraction of external ®nancing which
comes from intermediaries, (Figure 6.3a) and the `liability-intermediation
ratios of non-®nancial companies to banks', which measure the share of bank
®nancing in the total external ®nancing of ®rms (Figure 6.3b), and the ratios of
securitisation of corporate ®nancing (Figure 6.3c) differ greatly between
Germany and Great Britain, and are almost completely stable in these two
countries. For France, these ratios exhibit not only a great instability, but also a
tendency to move away from the German to the British model.
This may suf®ce to demonstrate that the general characterisations of the three
®nancial systems are empirically valid, that the overall ®nancial systems of
Germany and the United Kingdom are consistent con®gurations of the sub-
systems as complementary elements and that there are indeed considerable and,
at least in some cases, persistent differences between the ®nancial systems of the
major European countries. Having established this, I can now analyse what
consequences this might have for a common monetary policy within EMU.
4 Consequences for EMU: ®nancial systems' structures and

common monetary policy
Is there a common reaction of the different national systems to monetary

policy?
After having argued that the credit channel is probably relevant in Europe, too,
and that ®nancial systems in Europe differ in fundamental ways, I will now
a) Liability–intermediation ratios
a) Liability–intermediation of non-financial
ratios companies
of non-financial (NFCs)
Companies (NFCs) b) Liability–intermediation ratios of
of NFCs
banks to banks
Financial liabilitiesofNFCs
Financial Liabilities NFCstotoFS/
FS/ T
Total Financial
financial Liabilities
liabilities of
of NFCs
NFCs Financial liabilities of
Financial Liabilities NFCs
NFCstotobanks/
banks/ TTotal financialLiabilities
otal financial liabilitiesofofNFCs
NFCs
85 80
80 Germany 70
70 Germany
75 60
60
70
65 50
50
France
60 40
40
United Kingdom
55 30
30
50 20
France 20
45
United Kingdom
40
10
35 0
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
c)c)Liability–securitisation ratios
Liability–Securitisation of of
ratios non-financial companies
Non-financial companies d)
d) Liability
Liability [Asset]–intermediation
[Asset]–intermediation ratios
ratios of
of banks
banks
Securitised
itisation liabilities FinancialLiabilities
Financial liabilities[assets]
[assets]of of banks
banks to [from]
to [from] NBFIs/
NBFIs
Secur Liabilities of of NFCs
NFCs / Total
/ Total financial
financial liabilities
Liabilities of NFCs
of NFCs
Totalfinancial
Total financialliabilities
liabilities [assets]
[assets] of NFCs
of banks
80
United Kingdom 20 Liability–intermediation ratios
70
France
60 15
France
50 Germany
40 10
30 d United Kingdom
Germany 5
20
10
A-IRs
0 0
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
1981
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
Figure 6.3 Intermediation and securitisation ratios (in percent), 1981±96
offer a theoretically inspired speculation as to what main consequence for a

common monetary policy these differences entail. One could be inclined to
think that marked differences make it very dif®cult to design and to
implement a common monetary policy. Is this conclusion justi®ed or would
adopting this position ± which would, of course, also have signi®cant political
implications ± mean simply to jump to conclusions?
The transmission of monetary policy goes from the central bank (CB) to the
(other parts of the) ®nancial sector (FSec) and from there to the real economy
(RE), which reacts to monetary developments by changing investment and
possibly consumption decisions, which in turn determines aggregate income,
employment and the price level. The top line in Figure 6.4 (p. 228) shows the
elements of the transmission mechanism. A denotes the relationship between
CB and FSec while B denotes the relationship between FSec and RE.
Since I wish to focus on different ®nancial systems in a monetary union,
differences between countries at the level of CB are of no concern for my
mental experiment, as I assume that there is just one single CB. But ®nancial
systems may vary in four respects: (1) The way in which the ®nancial sector
(FSec) is in¯uenced by central bank policy measures (A) can, and probably does,
differ; and (2) the FSecs may be, and indeed are, dissimilar; and (3) the ways in
which the ®nancial sector changes credit supply to ®rms and households, and
in which the real sectors change their demand for ®nancial services, in
reactions to the changing conditions in the ®nancial sector (B) seem to be
different; and (4) important aspects of RE ± notably company ®nancing,
governance and strategies and structures ± are likely to differ. Or to put it more
succinctly: ®nancial and real sectors and the sensitivities of the ®nancial sector
to central bank policies, and of the real sectors to changing conditions in the
®nancial sector may differ between countries.
Comparing Germany and the United Kingdom, one can ask: in which
system does a central bank impulse of a given kind and size have a stronger effect
on the ®nancial sector (A)? And in which of the two countries does a given
amount of in¯uence of the central bank on the ®nancial sector have a stronger
effect on the aggregate demand of the real sectors (B) ± and ultimately on
income, employment and prices?
The answers depend on how one sees these relationships on the structures of
the ®nancial systems and in particular on the roles of the banking sector in the
respective ®nancial system. I use the concepts underlying the notion of a
credit channel. Here the immediate effect of central bank action is assumed to
be mainly an effect on the banks. Thus, the strength of the relation A between
the central bank and the ®nancial sector is likely to be stronger in a country in
which the relative importance of banks in the ®nancial sector is greater. This
should be the case in Germany because of the following main reason. The
German ®nancial sector consists mainly of universal banks which supply
nearly all kinds of ®nancial services under one roof.35 In contrast to the British
®nancial sector banks do not have any serious NBFI-competitors in
Germany.36 Thus the central bank can determine more directly the
re®nancing conditions of a much larger part of the ®nancial sector. I therefore
hypothesise that the ®nancial sector of Germany reacts more strongly to
monetary policy impulses than the British ®nancial sector.
As a second step, one has to ask how changing conditions in the entire
®nancial system ± and, in particular, in the banking sector ± affect decisions of
®rms and households. To provide a concrete example, let us examine the case
of a tightening of monetary policy.37 Would funding opportunities and
funding costs for the real sectors be restricted or raised more in Germany or in
the United Kingdom? I presume that the restrictive effect of a given in¯uence
of the central bank on the ®nancial sector ± and, in particular, on the situation
of the banks ± would be less in Germany than in the United Kingdom, and this
for three reasons.
The ®rst reason is that the relationship between banks on the one side and
®rms and households on the other is closer in Germany than in the United
Kingdom. Owing to their closer multifaceted and more long-term-oriented
relationships with their customers, German banks might be more hesitant to
tighten credit terms for their clients in order not to burden them too much and
not to disturb their longer-term investment and business strategies in which a

`housebank'38 has a lively interest. In a system like that of the United
Kingdom, in which arm's length banking prevails, banks and other ®nancial
intermediaries simply have less reason to refrain from passing on restrictive
monetary policy impulses to their customers; they are less committed to their
long-term strategies.39 An additional argument is that, compared with a
typical British bank, the typical German bank would have more information
on its customers, and would thus need to worry less about their net values
declining as a consequence of higher interest rates.40
The second reason is that credit terms differ systematically between
countries, and that these differences are in line with the general characteristics
of the respective ®nancial systems. In Germany, credit contracts have a longer
maturity on average, and interest rates are typically less adjustable than in the
United Kingdom. Thus, the ®nancial sector would simply ®nd it more dif®cult
to pass on rising interest rates. Furthermore, German accounting principles
allow ®rms to build up hidden reserves to a much greater extent. As a
consequence, the investment decisions of German ®rms should be less
sensitive to short-run changes in the costs and availability of external funds.
By contrast, UK accounting principles are such that ¯uctuations in the cash
¯ow have a greater in¯uence on investment, which in turn ampli®es the
effects of larger ¯uctuations in the cash ¯ow on aggregate demand and
income.
The third reason why non-®nancial companies in Germany do not react as
sensitively when the banks and the rest of the ®nancial sector ®nd themselves
in a more restrictive situation is to be found in the prevailing governance and
ownership patterns in both the ®nancial and the non-®nancial sectors. The
standard models in the macroeconomic literature about the reactions of banks
to a restrictive monetary policy of the central bank and about the reactions of
®rms' investment decisions to a tightening of credit terms are almost always
based on the assumption that all economic agents are strictly pro®t-oriented
and have a rather short-term perspective. This assumption is not equally valid
for German banks and corporations on the one side and UK banks and
corporations on the other. The bulk of the institutions comprising the German
banking system are not privately owned, pro®t-oriented joint stock corpora-
tions or limited liability companies. And even those big banks in Germany
which are joint stock corporations have a governance system which shields
their management almost perfectly from the performance pressure of their
shareholders, and thus their behaviour may not be too different from that of a
`Sparkasse' or a `Volksbank'. Thus the assumption that, as a general rule, banks
will react to central bank impulses in accordance with standard models, may
be wrong, at least for certain ®nancial systems. In Germany, for instance, banks
have a long-term interest in maintaining stable and `healthy' relations with
`healthy' clients, and it seems plausible to assume that they pursue this interest
by dampening monetary shocks, irrespective of whether these are intended by
the central bank or not. At the level of large companies, a strict and exclusive
short-term pro®t orientation41 is also the exception and not the rule in
Germany. The more moderate pro®t orientation of many large corporations
tends to make them ìgnore' changing ®nancial conditions so as to promote
their long-term strategies, and they probably are more prone to do this than
their counterparts in the United Kingdom. Here I would like to reiterate that
this behavioural pattern on the part of German banks and the large ®rms is not
necessarily a `problem' or a `¯aw'; rather, it is a feature that is consistent with
the general logic of the German ®nancial system.42
In the United Kingdom, both banks and corporations are more forced to
strive for pro®t and in so doing react to changing price signals. Both in their
relationships with clients and in the way they value outstanding securities,
privately owned and unambiguously pro®t-oriented banks and other ®nancial
intermediaries react more strongly to central bank activity; and private ®rms
which have to be extremely conscious of their stock price performance are
more strongly motivated ± indeed, forced ± to react more to price signals, as it
is to be expected in a more market-oriented system.
Thus one can conclude that the two systems also have different reaction
functions with respect to the relationship between the ®nancial sector and the
non-®nancial sectors (B). The British system reacts more, and the German
system reacts less, to changes in the situation of the respective ®nancial
sectors. Figure 6.4 summarises the two parts of the argument presented so far.
As can be seen, the German system combines a stronger in¯uence of the
central bank on the situation of the ®nancial sector (A, with four stars in the
second line of Figure 6.4) and a weaker in¯uence of the ®nancial sector on the
real economy (B, with two stars). The situation in the United Kingdom is
the reverse. The dominance of the central bank over the ®nancial system is
weaker (A, with two stars in the third line) and that of the ®nancial sector over
the real sector is stronger (B, with four stars).
It would, of course, be naive to quantify the strength of these effects solely on
the basis of the arguments which I have provided. Accordingly, all that I would
like to do here is to present the idea that the relative strengths of the two
relationships A and B are inversely related. That this should be so, is not by
coincidence. Instead, it follows from the systemic features of the ± broadly
de®ned ± ®nancial systems, which are consistent in each of these two countries.
If one carried the exercise of `counting stars' to an extreme, one would ®nd
that the net effect of monetary policy may be the same in both countries under
consideration here. But I do not mean to imply that the two differences
between the partial channels A and B in the two countries cancel
out completely. I chose to use stars, and not numbers, because the latter might
have suggested a greater degree of precision, perhaps leading to
misunderstandings.
However, it does seem safe to assume that the differences between the
stylised ®nancial systems have offsetting effects with respect to A and B, and in
A B
CB FSec RE
Credit Germany
channel
United Kingdom
Interest rate Germany

channel
United Kingdom
Figure 6.4 Monetary policy effectiveness in Germany and the United Kingdom
itself this may be an important factor for assessing the potential for a common
monetary policy in an enlarged Èuroland'. The two ®nancial systems may
differ in a fundamental way, and yet the net effectiveness of monetary policy
may not be all that different. If this is news, it is good news for the proponents
of a monetary union in Europe and its future enlargement.
The argument presented here in a very non-technical manner implicitly
draws on the credit channel literature. It could be rendered more precise and
more technical by making this basis explicit and spelling out in some detail
how the differences between the ®nancial systems determine the strength of
the relationships A and B in the two countries. Not unexpectedly, this exercise
would demonstrate that the effects of institutional features on the transmis-
sion mechanism are complex, dif®cult to aggregate and in some cases
ambiguous. Nevertheless, the basic thrust of the argument would not change;
the overall effects of the working of the bank lending and the balance sheet
channels are in line with my intuitive argument.43 But even a more technical
approach would not solve one important aggregation problem: How can one
àdd' the different effects of the links A and B given that the relevance of the
link B depends on how strong the link A is? It seems that this question requires
much more work and in-depth econometric studies. I would be delighted if the
brief non-technical discussion I have presented here were to be the inspiration
for such a research project.
Having admitted that my attribution of stars to the partial channels A and B
in the two stylised ®nancial systems owes a lot to the credit channel literature,
I should brie¯y add a conjecture concerning whether the result is stable with
respect to the theoretical basis. The general thrust of the credit channel
argument is that the possible effects of monetary shocks, including those
engineered by a central bank, are probably greater than those which the
literature based on the interest rate channel takes into account. If one retains
the notion of the ®nancial system as a system of complementary elements,
and the premise that Germany and the United Kingdom have consistent but
differing ®nancial systems, and combines them with the assumptions of the
interest rate channel, then the effects for both partial channels would be
weaker: the effect of the central bank on the banking system (A), which could
be approximated by the multiplier, would be less pronounced in both
countries and the difference between the effect in Germany and the effect in
the United Kingdom would still be about the same. The effect of a change in
the situation of the banks and other ®nancial institutions on the real economy
(B) would probably also be weaker in general ± and at least some of the
arguments which would suggest that in the United Kingdom this effect is
stronger would also still apply. Thus, the main result ± namely, that the two
parts, A and B, of the transmission mechanisms tend to exhibit offsetting
differences between the two countries ± would also be valid if one disregarded
the speci®c aspects of the credit channel. This is shown by the stars bracketed
on the lower lines in Figure 6.4.
Are the different reaction functions likely to be stable?

I would like to conclude the chapter with some brief remarks about dynamic
issues. Financial systems and their sub-systems are not immutable. European
integration, regulatory convergence and, last but not least, the common
currency have already brought important changes, and are likely to lead to
more change in the future. What does this imply for the argument presented
in the preceding section?
I do not believe that the introduction of the Euro and even the
enlargement to include the United Kingdom will change the basic
structures of the ®nancial systems of the participating countries in any
fundamental way. Money and central banks are not core elements of the
concept of ®nancial systems on which the main proposition in this chapter
is based. Different countries can have different ®nancial systems in the
sense of consistent sets of complementary elements and sub-systems even
though they have a common currency and thus necessarily an almost
identical monetary policy. This suggests that the introduction of the Euro
per se will also not invalidate my proposition that some ®nancial systems in
Europe are fundamentally different and that a common monetary policy
may nevertheless be feasible.
The common currency and especially the enlargement to include the
United Kingdom will, however, magnify all of the forces which are
currently working to transform the ®nancial systems, and speed up ongoing
processes of change. This raises the question of how ®nancial systems
change in general and what this implies for the common monetary policy
in Èuroland'. Owing to space constraints, I will have to restrict myself to a
few brief remarks.
The `fact' that ®nancial systems have the property that every element of a
®nancial system is functionally related to many others and that a consistent
®nancial system constitutes a local optimum is relevant here. It implies that
partial reforms are not likely to be sustainable, and this in turn implies that
®nancial systems are not likely to change and to converge gradually.
But if there is no gradual convergence, what else could happen? If, for
instance, political forces or dynamic entrepreneurs in the ®nancial service
industry succeed in introducing elements of one system ± for instance, the
active takeover market of the British system ± into the other ± for example, that
of Germany ± the immediate result would be an inconsistent `non-system'. If
this happens, there will be pressure to restore consistency. A `restauration' of
the old system is one way of achieving consistency. In this case the
ìnnovation' is rejected. This is one possible course of events.
The other possibility is that the forces which make changes appear
desirable are strong and the forces of `restauration' weak and that the
formerly consistent system of a given country undergoes not only a series
of partial changes, but will in fact experience a `Gestalt switch' from one
type to the other. The resulting new con®guration of the elements of the
system may be better or worse; one local optimum is replaced by another
one.44
In both cases ± that is, in the case of partial reforms which only lead to a
`restauration' and in the case of a complete transformation of the system's
architecture ± the task of monetary policy makers will be extremely dif®cult
because the transmission mechanism will have become unstable. In such a
situation the European Central Bank would scarcely be able to predict the
overall effects of its policy on the economy of Èuroland' and would
therefore ®nd it dif®cult to determine the precise strength of the monetary
impulses which it should provide. There would be less of a chance of such a
destabilisation and loss of orientation of monetary policy makers occurring
in Europe as a consequence of changes in the national ®nancial systems if
different national currencies and monetary policies had been retained and
not replaced by a uni®ed currency regime; and under the old system of
national monetary policies, such a destabilisation would have less serious
consequences. I want to conclude by expressing my belief that this
increased risk of instability and disorientation constitutes a bigger problem
for a common currency than the need to design and implement a common
monetary policy for different, but essentially stable, ®nancial systems, on
which the existing literature focuses.
Appendix
Table 6.A1 Financial system indicators: ®nancial sectors
Germany United France

Kingdom
(a) Capital markets

Number of domestic listed shares (1997) 699 2465 683
Market capitalisation of shares related to
GDP (1996) (%) 22 136 30
New listings between 1995 and 1997 60 765 109
Certi®cates of deposit as a percentage
of GDP (1997) 0.32 10 18.1
(b) Banks
Market share as percentage
of deposits by the non-bank public (1990)
Commercial banksa 39.7 56.7 53.6
Savings banks 36.1 3.4
Co-operative and rural banksb 19.6 42.7 30.1
Five-®rm concentration ratio of commercial
banks (1995) 17 57 47
Interest margin (1995) 2.66 2.21 1.66
Pro®t before tax (average 1990±4
as percentage of balance sheet total) 0.55 0.66 0.06
(c) Non-bank ®nancial intermediaries
Private pension ®nancing as percentage
of GDP (1996) 53 75.6 4.5
Pension funds portfolio composition
(1996)(%)
equities 8 78 14
bonds 74 14 38
property 7 5 8
liquid assets 12 4 40
Notes:
a
For UK authorised banks.
b
UK building societies.
Sources: Davis (1998); Wymeersch (1998); European Central Bank (1999).
Table 6.A2 Financial system indicators: corporate governance

Kingdom
(a) Takeovers (1988±96) 4 1190 155

(b) CEO compensation structure (1998)
Total remuneration (average) in (US $) 398 430 645 540 520 389
Variable bonus as a percentage
of annual basic compensation 27 22 19
Options/long-term incentive plans as a
percentage of annual basic compensation 2 38 30
(c) Ownership concentration of listed
companies (1994±5)
Main shareholder > 50% 68 7 37
25±50% 21 12 32
< 25% 11 81 31
(d) Share ownership structure (1995) (%)
Private households/individuals 14.6 20.3 19.4
Public sector 4.3 0.8 3.4
Insurance companies 12.4 21.9 1.9
Pension/Investment-funds 7.6 36.6 2.0
Banks 10.3 0.4 4.0
Enterprises/commercial corporations 42.1 1.1 58.0
Rest of the world 8.7 16.3 11.2
Sources: Wymeersch (1998); Towers Perrin's 1998 Worldwide Total Remuneration Report.
Table 6.A3 Financial system indicators: ®nancial patterns

Kingdom
(a) Corporations
Dividends/cash ¯ow ratio for listed
companies (1994) (%) l2.7 l6.67 9.46
Corporate loans collateralised by real
estate as percentage of total corporate loan (1993) 36 59 4l
Short-term corporate credit
as percentage of total corporate loan (1993) 22 50 27
Long-term corporate credit
as percentage of total corporate loan (1993) 78 50 73
(b) Households
Mortgage credit interest adjustment (%)
Fixed 20 0 80
Renegotiable 40 30 0
Variable 0 0 20
Reviewable 40 70 0
Typical loan±value ratio 60±80 90±95 70±80
Notes:
Fixed: rate ®xed until ®nal maturity.
Renegotiable: rate not ®xed over entire term, but more than one year
Variable: rate adjustable according to index reference rate.
Reviewable: rate adjustable at discretion of lender.
Sources: Maclennan, Muellbauer and Stephen (1998); Borio (1995); La Porta et al. (1999)
Notes
1. The author wishes to thank Falko Fecht, Andreas Hackethal and Adalbert Winkler,
and in particular Marcel Tyrell, for extremely valuable advice and research support.
Financial support of the Deutsche Forschungsgemeinschaft is gratefully acknowl-
edged.
2. There is a growing body of econometric literature on differences regarding the
monetary transmission mechanisms in European countries (see for instance Bank
for International Settlements, 1995; Dornbusch, Favero and Giovazzi, 1998; Britton
and Whitley 1997; Ramaswamy and Sloek, 1997; Giovanetti and Marimon 1998).
To me this literature, which is partly surveyed in Dornbusch, Favero and Giovazzi
(1998), does not provide unambiguous evidence of great differences in the
transmission mechanisms. However, to the extent that these papers discuss
differences between national ®nancial systems at all, they do not go very far in this
respect. But see Kashyap and Stein (1997a), who follow a similar approach to the
one in the present chapter.
3. See Schmidt and Tyrell (1997) for this terminological distinction.
4. See Schmidt (1990).
5. See for instance the articles by Bernanke and Gertler (1995); Meltzer (1995);
Mishkin (1995) and Taylor (1995) in the Journal of Economic Perspectives and
symposium on monetary transmission mechanism, and Goodhart (1989);
Cecchetti (1995) and Illing (1997, pp. 145 ff.) for overviews.
6. Especially to households' decisions concerning the acquisition of homes and of
consumer durables.
7. See Borio (1997) for an exhaustive analysis of different monetary policy procedures
and their recent convergence.
8. See Meltzer (1995).
9. See Bernanke and Gertler (1995) for this argument.
10. For surveys of this transmission channel with empirical results for the United
States, see also Bernanke, Gertler and Gilchrist (1996) and Kashyap and Stein
(1997b).
11. See Bernanke, Gertler and Gilchrist (1996).
12. For an overview see Freixas and Rochet (1997). More recent contributions include
Rajan (1996) and Kashyap, Rajan and Stein (1999). The quotation paraphrases the
title of an in¯uential article by James (1987).
13. So far, and according to my knowledge, the credit channel has been the subject of
only a few empirical investigations in Germany. Sto È û (1996) and Guender and
Moersch (1997) come to a negative conclusion concerning the importance of the
bank lending channel in Germany. Worms (1997) ®nds some positive evidence
with respect to the balance sheet channel. Kuppers (1998a, 1998b) forcefully
criticises the results of Guender, Moersch and Sto È û and ®nds strong support for a
credit channel in his own empirical study. For the United Kingdom Dale and
Haldane (1995) and Ganley and Salmon (1997) show some importance of the
credit channel. More recent research on the credit channel in France includes Goux
(1996), Candelon and Cudeville (1996) and Payelle (1996). Their results are
somewhat ambiguous, but support the assumption that the credit channel is
relevant in France too.
14. The following discussion is based on Hackethal and Schmidt (2000). See also
Milgrom and Roberts (1995); Hackethal and Tyrell (1998) and Aoki (1999).
15. For this de®nition see also Milgrom and Roberts (1995). The mathematics behind
the concept of complementarity are surveyed by Topkis (1998).
16. See Rybczinski (1984) and Berglof (1990) for this classi®cation with respect to the
®rst and second sub-system.
17. These dichotomies and the way in which they are related, are discussed in
Hackethal and Schmidt (2000).
18. See Franks and Mayer (1994); Schmidt (1997b) and Tirole (1999).
19. See Boot and Macey (1998), Hackethal and Schmidt (2000); Aoki (1999).
20. The link between bank-oriented ®nancing, insider-controlled governance and ®rm-
speci®c human capital is more deeply analysed in Hackethal and Tyrell (1998) and
Berkovitch and Israel (1998). The correspondence to the business systems is
discussed by Milgrom and Roberts (1995); Aoki (1999) and Hackethal and Schmidt
(2000). See also Mayer (1998).
21. See Williamson (1988).
22. See also Goodhart (1993).
23. See the empirical results in European Central Bank (1999) and Davis (1998). Some
indicators are presented in the Appendix.
24. See Schmidt and Tyrell (1997) and Prigge (1998).
25. For the United Kingdom we do not know enough about (4), but see Hackethal and
Schmidt (2000) for ®rst results and some rather speculative conclusions.
26. See Goergen and Renneboog (1998) and Franks, Mayer and Renneboog (1998).
27. See Charkham (1994); Franks and Mayer (1997) and Wymeersch (1998).
28. For a similar characterisation of the British system, see also Prevezer and Ricketts
(1994).
29. See Schmidt (1997a) for some details. The recent book by Plihon (1998, p. 79), is
among the numerous supporting French references which one could quote here.
30. For a similar conclusion see OECD (1995).
31. For instance, this can be concluded from the more active market for corporate
control in recent years (Wymeersch, 1998). On the other hand, ownership
concentration and voting power in French public corporations indicates an insider-
control system (Bloch and Kremp, 1998).
32. See Mayer (1988, 1990) and Corbett and Jenkinson (1996, 1997) for international
comparisons of ®nancing patterns and Edwards and Fischer (1994) for a study of
Germany and Bertero (1994) for France.
33. In the work of Mayer and his followers, one can ®nd another distinction. It is the
distinction between net and gross ®gures, which concerns a different aspect from
the one under discussion here. Their 'gross ®gure' are calculated after the
aggregation which is identi®ed in Hackethal (1999) and Hackethal and Schmidt
(1999) as the cause of the bias.
34. See Schmidt, Hackethal and Tyrell (1999).
35. See for instance the detailed descriptions of the British and German ®nancial
sectors in Saunders and Walter (1996).
36. The predominance of the universal banks in Germany may be due to regulatory
conditions partly resulting from the monetary policy strategy of the Deutsche
Bundesbank, which heavily depended on a stable money demand. See for instance
the Deutsche Bundesbank (1995).
37. This is more than a way of making the discussion more concrete. It might well be
that the arguments which follow, only apply to the special case of a restrictive
monetary policy.
38. This re¯ects that there is some truth to the belief that the old system of
'housebanks' still prevails. This has been vigorously challenged in a well known
book by Edwards and Fischer (1994). But note that the empirical basis of their
attack on this presumed myth is an empirical analysis of ®nancing patterns using

the methodology of Colin Mayer (1988), which was discussed as yielding unreliable
empirical results on p. 222. Studies by Elsas and Krahnen (1998) and Harhoff and

Korting (1998) indicate that banks which perceive themselves as the 'housebank' of
a given customer ®rm, and which are also perceived as such by the customer ®rm
itself, behave systematically differently, that is in a more cooperative and more
long-term oriented manner. Thus relationship banking and housebanks still seems
to be real factors.
39. Furthermore, a study of the monetary transmission mechanism in Germany with
segregated bank balance sheet variables by Kuppers (1998a) demonstrates the
relevance of the 'housebank relationship'. His results show that, in contrast to
empirical ®ndings for the United States by Kashyap and Stein (1997b), credit terms
and loan volumes of smaller German banks, namely the Savings Banks and the
Cooperative Banks, react to a monetary tightening to a much smaller extend than
the German 'Grossbanken'. In doing so, they isolate their customers from

monetary shocks. Kuppers argues that Savings Banks and Cooperative Banks are
those credit institutions in Germany which conform most to the model of a
housebank with respect to a large number of their customers.
40. That this more ef®cient way of gathering and processing information by banks can
also result in more ef®cient investment decisions by ®rms, is theoretically shown in
Dewatripont and Maskin (1995) and von Thadden (1995).
41. A genuine pro®t orientation is typically an orientation towards short-term pro®ts
because otherwise it would be dif®cult to make this orientation operational. Long-
term pro®t orientation would be indistinguishable from an orientation towards the
maximisation of total value or growth potential or 'strategic advantage'; see
Schmidt and Massmann (1999).
42. This argument re¯ects the effect of intergenerational risk smoothing through the
®nancial system which is discussed on a theoretical level by Allen and Gale (1997).
These authors show that the limited ± or merely long-term ± pro®t orientation of
the banking sector can lead to welfare gains.
43. I am extremely grateful to Marcel Tyrell and Falko Fecht for having helped me to
arrive at this result, which is not elaborated in the present chapter because of space
limitations.
44. In Schmidt and Spindler (1999) it is shown that, and why, there is a possibility that
in the 'competition' between different (consistent) national systems of corporate
governance the one which is less ef®cient under stable conditions may be
universally adopted.
References
Allen, F. and D. Gale (1997), `Financial Markets, Intermediaries and Intertemporal
Smoothing', Journal of Political Economy, 105, 23±46.
Aoki, M. (1999), 'An Information Theoretic Approach to Comparative Corporate
Governance', Department of Economics, Stanford University, Working Paper 99±04.
Bank for International Settlements (1995), Financial Structure and the Monetary Policy
Transmission Mechanism, Basle: Bank for International Settlements.
E. (1990), `Capital Structure as a Mechanism of Control: A Comparison of
Berglof,
Financial Systems', in M. Aoki, B. Gustafsson and O. E. Williamson (eds), The Firm as a
Nexus of Treaties, London: Sage, 237±62.
Berkovitch, E. and R. Israel (1998), 'Optimal Bankruptcy Laws Across Different Economic
Systems', University of Michigan, Working Paper.
Bernanke B. and M. Gertler (1995), Ìnside the Black Box: The Credit Channel of
Monetary Policy Transmission', Journal of Economic Perspectives, 9, (4) 27±48.
Bernanke, B., M. Gertler and S. Gilchrist (1996), `The Financial Accelerator and the Flight
to Quality', Review of Economics and Statistics, 78, 1±15.
Bertero, E. (1994), `The Banking System, Financial Markets, and Capital Structure: Some
New Evidence from France', Oxford Review of Economic Policy, 10, (4),68±78.
Bloch, L. and E. Kremp (1998), Ownership and Voting Power, Louvain, European Corporate
Governance Network.
Boot, A. W. A. and J. R. Macey (1998), 'Objectivity, Control and Adaptability in
Corporate Governance', Tinbergen Institute, Discussion Paper, 98±064.
Borio C. E. V. (1995), `The Structure of Credit to the Non-Government Sector and the
Transmission Mechanism of Monetary Policy: A Cross-Country Comparison', in Bank
for International Settlements (ed.), Financial Structure and the Monetary Policy
Transmission Mechanism, Basle: BIS, March, 59±105.
ÐÐÐÐ (1997), `The Implementation of Monetary Policy in Industrial Countries: A
Survey', BIS Economic Paper, 47.
Britton, E. and J. Whitley (1997), `Comparing the Monetary Transmission Mechanism in
France, Germany and the UK: Some Issues and Results', Bank of England Quarterly
Bulletin, May,152±62.
Candelon, B. and E. Cudeville (1996), `Politique monetaire
et canal de credit: une

estimation empirique sur l'economie
francËaise', Revue d'economie politique, 104,
781±807.
Cecchetti, S. G. (1995), `Distinguishing Theories of the Monetary Transmission
Mechanism', Federal Reserve Bank of St. Louis Review, May±June, 83±97.
Charkham, J. P. (1994), Keeping Good Company: A Study of Corporate Governance in Five
Countries, Oxford, Oxford University Press.
Corbett, J. and T. Jenkinson (1996), `The Financing of Industry, 1970±1989: An
International Comparison', Journal of the Japanese and International Economies, 10,
71±96.
ÐÐÐÐ (1997), `How is Investment Financed? A Study of Germany, Japan, The United
Kingdom and the United States', The Manchester School Supplement, 69±93.
Dale, S. and A. G. Haldane (1995), Ìnterest Rates and the Channels of Monetary
Transmission: Some Sectoral Estimates', European Economic Review, 39, 1611±1622.
Davis E. P. (1998), `Pension Fund Reforms and European Financial Markets', Financial
Markets Group, London School of Economics, Special Paper 107.
Deutsche Bundesbank (1995), `Verbriefungstendenzen im deutschen Finanzsystem und
ihre geldpolitische Bedeutung', Monatsbericht , April, 19±34.
Dewatripont, M. and E. Maskin (1995), `Credit and Ef®ciency in Centralized and
Decentralized Economies', Review of Economic Studies, 62, 541±55.
Dornbusch, R., C. Favero and F. Giavazzi (1998), Ìmmediate Challenges for the
European Central Bank', Economic Policy, 26, 15±64 (including discussion).
Edwards, J. and K. Fischer (1994), Banks, Finance and Investment in Germany, Cambridge,
Cambridge University Press.
Elsas, R. and J. P. Krahnen (1998), Ìs Relationship Lending Special? Evidence from
Credit-File Data in Germany', Journal of Banking and Finance, 22, Special Issue on Credit
Risk Assessment and Relationship Lending, 1283±1316.
European Central Bank (1999), Possible Effects of EMU on the EU Banking Systems in the
Medium to Long Term, Frankfurt: European Central Bank.
Franks, J. and C. Mayer (1994), 'Corporate Control: A Comparison of Insider and
Outsider Systems', London Business School, Working Paper.
Franks, J. and C. Mayer (1997): Corporate Ownership and Control in the United Kingdom,
Germany and France, Journal of Applied Corporate Finance, 9, 30±45.
Franks, J., C. Mayer and L. Renneboog (1998), 'Who Disciplines Bad Management',
London Business School, Working Paper.
Freixas, X. and J. C. Rochet (1997), Microeconomics of Banking , Cambridge, Mass: The
MIT Press.
Ganley, J. and C. Salmon (1997), 'The Industrial Impact of Monetary Policy Shocks:
Some Stylised Facts', Bank of England Working Paper 68.
Giovannetti, G. and R. Marimon (1998), 'An EMU with Different Transmission
Mechanisms', CEPR Working Paper 2016.
Goodhart, C. A. E. (1989), `The Conduct of Monetary Policy', Economic Journal, 99, 293±346.
ÐÐÐÐ (1993), `Banks and the Control of Corporations', The Cyprus Journal of
Economics, 6 (2), 91±108.
Goergen, M. and L. Renneboog (1998), Strong Managers and Passive Institutional Investors
in the UK, Louvain, European Corporate Governance Network.

Goux, J.-F. (1996), `Le canal etroit
du credit
en France', Revue d'economie politique, 104,
655±681.
Guender, A. and M. Moersch (1997), Òn the Existence of a Credit Channel of Monetary
Policy in Germany', Kredit und Kapital, 30, 173±185.
Hackethal, A. (1999), Unternehmens®nanzierung in Deutschland, USA und Japan, PhD
dissertation, University of Frankfurt (fortcoming), Frankfurt, Lang.
Hackethal, A. and R. H. Schmidt (1999), 'Financing Patterns: Measurement Concepts
and Empirical Results', University of Frankfurt,Working Paper Series: Finance and
Accounting, 33.
ÐÐÐÐ (2000), Komplementarita t und Finanzsystem, Kredit und Kapital, Beiheft 15
(inprint).
Hackethal, A. and M. Tyrell (1998), 'Complementarity and Financial Systems ± A
Theoretical Approach', University of Frankfurt, Working Paper Series: Finance and
Accounting, 11.

Harhoff, D. and T. Korting (1998), `Lending Relationships in Germany ± Empirical
Evidence from Survey Data', Journal of Banking and Finance, 22, Special Issue on Credit
Risk Assessment and Relationship Lending, 1317±53.

Illing, G. (1997), Theorie der Geldpolitik: Eine spieltheoretische Einfuhrung, Berlin: Springer.
James, C. (1987), `Some Evidence on the Uniqueness of Bank Loans', Journal of Financial
Economics, 19, (2), 217±35.
Kashyap, A. K. and J. C. Stein (1997a), `The Role of Banks in Monetary Policy: A Survey
with Implications for the European Monetary Union', Economic Perspectives, Federal
Reserve Bank of Chicago, 22, (5), 2±18.
ÐÐÐÐ (1997b), 'What Do a Million Banks Have to Say About the Transmission of
Monetary Policy?', NBER Working Paper, 6056.
Kashyap, A. K. , R. Rajan and J. C. Stein (1999), 'Banks as Liquidity Providers: An Explanation
for the Co-Existence of Lending and Deposit-Taking', NBER Working Paper, 6962.

Kuppers, M. (1998a), 'Curtailing the Black Box: German Banking Groups in the
Transmission of Monetary Policy', University of Cologne, Staatswissenschaftliches
Seminar Discussion Paper Series, 17.
ÐÐÐÐ (1998b), 'On the Existence of a Credit Channel of Monetary Policy in
Germany: A Comment', University of Cologne, Staatswissenschaftliches Seminar
Discussion Paper Series 19.
La Porta, R., F. Lopez-de-Silvanes, A. Shleifer and R. W. Vishny (1999), 'Agency Problems
and Dividend Policies Around the World', Harvard University, Working Paper.
MacLennan, D., J. Muellbauer and M. Stephens (1998), Àsymmetries in Housing and

Financial Market Institutions and EMU', Oxford Review of Economic Policy, 14, (3), 54±80.
Mayer, C. (1988), `Financial Systems and Corporate Governance: A Review of the
International Evidence', Journal of Institutional and Theoretical Economics, 154, 144±165.
ÐÐÐÐ (1990), `Financial Systems, Corporate Finance, and Economic Development',
in R. G. Hubbard (ed.), Asymmetric Information, Corporate Finance and Investment,
Chicago, Chicago University Press, 307±332.
ÐÐÐÐ (1998), `New Issues in Corporate Finance', European Economic Review, 32,
1167±88.
Meltzer, A. H. (1995), `Monetary, Credit and (Other) Transmission Processes: A
Monetarist Perspective', Journal of Economic Perspectives, 9, (4), 50±72.
Milgrom, P. and J. Roberts (1995), `Complementarities and Fit ± Strategy, Structure, and
Organizational Change in Manufacturing', Journal of Accounting and Economics, 19,
179±208.
Mishkin, F. S. (1995), `Symposium on the Monetary Transmission Mechanism', Journal
of Economic Perspectives, 9, (4), 3±10.
Organisation for Economic Co-operation and Development (OECD) (1995), National
Systems for Financing Innovations, Paris, OECD.
Payelle, N. (1996), `Conditions de foncionnement du canal du cre dit et gestion des bilans
bancaires: un approche empirique dans le cas francËais', Revue d'e conomie politique, 104,
683±703.
Plihon, D. (1998), Les Banques ± Nouveaux Enjeux, Nouvelles Strate gies, Paris, La
Documentation FrancËaise.
Prevezer, M. and M. Ricketts (1994), 'Corporate Governance: The UK Compared with
Germany and Japan', in N. Dimsdale and M. Prevezer (eds), Capital Markets and
Corporate Governance, Oxford, Clarendon Press, 237±56.
Prigge, S. (1998), À Survey of German Corporate Governance', in K. Hopt et al. (eds),
Comparative Corporate Governance ± The State of the Art and Emerging Research , Oxford,
Clarendon Press, 943±1044.
Rajan, R. (1996), `Why Banks Have a Future: Toward a New Theory of Commercial
Banking', Journal of Applied Corporate Finance, 9, 114±35.
Ramaswamy, R. and T. Sloek (1997), 'The Real Effects of Monetary Policy in the European
Union: What Are the Differences?', IMF Working Paper 97±160.
Rybczinski, T. (1984), Ìndustrial Financial Systems in Europe, US and Japan', Journal of
Economic Behaviour and Organization, 5, 275±286.
Saunders, A. and I. Walter (1996), Universal Banking: Financial System Design Reconsidered,
Chicago, Irwin.

Schmidt, R. H. (1990), Ìnformationsokonomie
und Preisentwicklung an Finanzmarkten:
Abschied von neoklassischen Optimierungsvorstellungen?', in W. FilcË and C. Kohler
(eds), Kooperation, Autonomie und Devisenmarkt , Berlin: Duncker & Humblot, 13±34.
ÐÐÐÐ (1997a), `Comparing the French and German Financial System', in H. Kossbiel

(ed.), Internationale und Europaische Finanzsysteme, Frankfurt: Schulz-Kirchner, 9±33.
ÐÐÐÐ (1997b), `Corporate Governance: The Role of Other Constituencies', in
A. Pezard and J.-M. Thiveaud (eds), Corporate Governance: Les Perspectives Internatio-
nales, Paris, Montchre tien, 61±74.
Schmidt, R. H., A. Hackethal and M. Tyrell, `Disintermediation and the Role of Banks in
Europe: An International Comparison', Journal of Financial Intermediation, 8, 36±67.
Schmidt, R. H. and J. Massmann, J. (1999), `Drei Miûversta ndnisse zum Thema
``Shareholder Value''', in B. Kumar et al. (eds), Unternehmensethik und die Transformation
des Wettbewerbs, Stuttgart: SchaÈffer-Poeschel, 125±7.
Schmidt, R. H. and G. Spindler (1999), 'Path Dependence, Corporate Governance and

Complementarity ± A Comment on Bebchuk and Roe', University of Frankfurt,
Working Paper Series: Finance and Accounting, 27.
Schmidt, R. H. and M. Tyrell (1997), `Financial Systems, Corporate Finance and
Corporate Governance', European Financial Management, 3, 159±187.
E. (1996), 'Die Finanzierungsstruktur der Unternehmen und deren Reaktionen auf
Stoû,
moneta re Impulse ± Eine Analyse anhand der Unternehmensbilanzstatistik der
Deutschen Bundesbank', Deutsche Bundesbank, Discussion Paper 9/96.
Taylor, J. B. (1995), `The Monetary Transmission Mechanism: An Empirical Framework',
Journal of Economic Perspectives, 9, (4), 11±26.
Tirole, J. (1999), 'Corporate Governance', University of Toulouse, Working Paper.
Tobin, J. (1969), À General Equilibrium Approach to Monetary Policy', Journal of Money,
Credit, and Banking, 1, 15±29.
Topkis, D. (1998), Supermodularity and Complementarity, Princeton, Princeton University
Press.
Von Thadden, E.-L. (1995), `Long-term Contracts, Short-term Investment, and
Monitoring', Review of Economic Studies, 62, 557±575.
Williamson, O. E. (1988), `The Logic of Economic Organization', Journal of Law,
Economics, and Organization, 4, 65±93.
Worms, A. (1997), Bankkredite an Unternehmen und ihre Rolle in der geldpolitischen
Transmission in Deutschland, Frankfurt: Lang.
Wymeersch, E. (1998), 'A Status Report on Corporate Governance Rules and Practices in
Some Continental European States', in K. Hopt et al. (eds), Comparative Corporate
Governance ± The State of the Art and Emerging Research, Oxford: Clarendon Press,
1045±1210.
Discussion
Charles A. E. Goodhart
Schmidt's main result, as you will have read, is that ®nancial systems consist of
a number of complementary elements, with both Germany and the United
Kingdom having consistent but differing ± even polar opposite ± systems. I
accept this, and I am strongly predisposed towards accepting the counter-
attack launched against the Mayer position by Schmidt and his co-author and
student, Hackethal ± which is that the appropriate comparison is on gross, not
net, ®nancial ¯ows, and that on this basis the differentiation between the
German±Japanese and Anglo±Saxon ®nancial systems does show up reason-
ably clearly, (whereas it does not do so on a net basis).
I was also very interested in Schmidt's Figure 6.3 (p. 224) showing that the
French ®nancial system of intermediation was in several respects in the
process of moving from the German towards the British camp, as they had
presumably surmised in the Banque National de Paris (when launching a
hostile takeover bid). Indeed I had very little to criticise in sections on the
classi®cation of the various channels of monetary effects or on the general
characterisations of the ®nancial systems in these three countries.
But my role as a discussant is to comment critically and my particular
interest is on the effect of monetary policy on the real economies in Germany
and the UK. For this purpose I want to focus on Schmidt's Figure 6.4 (p. 228)
(see Figure D6.1).
Let me start with the credit channel. Here Schmidt claims that the effect on the
®nancial sector is larger in Germany than in the United Kingdom, largely because
the banks play a more central role in Germany, and I am prepared to accept that.
He also claims that the credit channel has a greater overall effect than the interest
rate channel and I again agree that that certainly can be the case, (consider, for
example, current Japanese problems), but I do not feel that this is necessarily so.
The credit channel, in my view, does not operate in a strictly linear fashion. In
some cases ± for example, where bank capital is already very plentiful ± it may be
weak, almost non-existent. In other circumstances it may, indeed, be ferocious.
So my classi®cation of the credit channel is as shown Figure D6.1b, distinguishing
between the weak and the (bracketed) strong cases.
241
242 Charles A. E. Goodhart
My main disagreement with Schmidt relates, however, to the interest rate

channel. With the United Kingdom being more market-oriented, and typically
with more short-term, impersonal arms-length contracts, I would expect both
links in the chain between monetary policy and the real economy to be
signi®cantly stronger in the UK than in Germany as shown in Figure D6.1b.
So my general assessment is that unless the monetary authorities press hard
enough to instigate a severe credit crunch ± and I rather doubt whether there
are any clear examples of that in Germany in recent years ± monetary policy
measures would otherwise bite quicker and harder in the United Kingdom
than in Germany. Schmidt argues that the asymmetries balance out, so that
there would be no problem for overall euro monetary management when, and
if, the United Kingdom joins.
Estimates from various models, VARs, structural VARs, structural models,
multicountry models, etc. of the comparative effects of monetary policy on
activity and in¯ation in the main European countries have come up with a
somewhat bewildering range of answers. Stephen Cecchetti in Chapter 5 in this
volume quotes work by Ehrmann, as well as his own prior work, indicating that
the effects on activity (and to a less marked extent on in¯ation), are less marked in
the Anglo±Saxon and Scandinavian countries than in France and Germany. I
recall other estimates that reverse this ranking, and certainly my own priors are
that monetary policy should bite harder in the Anglo±Saxon countries. And there
are yet other models that ®nd, or assume, no difference.
Be that as it may, my own position on this, that pure interest rate policy
bites harder in the United Kingdom, leads towards a further problem of
consistency with the evidence. If monetary policy does, indeed, bite harder in
the United Kingdom, then one would on the face of it expect interest rates to
have needed to be much less variable in the United Kingdom. Instead they
have been much more variable (see Figure D6.2). But these are nominal
interest rates. One answer is that German policy has been better in controlling
in¯ation, so that UK nominal rates have had to ¯uctuate more to offset the
larger variations in in¯ation; and there is some truth in that.
But even so, the variance of real interest rates has also, been somewhat
greater in the UK than in Germany (Figure D6.3). Why might this have been
so, if in the United Kingdom monetary policy does bite harder? Possible
reasons include:
(1) Greater shocks in United Kingdom;

(2) More nominal, fewer real, rigidities in Germany than in United Kingdom
± i.e. a ¯atter Phillips curve;
(3) Mismeasurement of the relevant index for in¯ation; and importance of
asset (housing) prices;
(4) More policy errors, and abrupt regime changes, in the United Kingdom;
Thus the in the last two decades has suffered far more variability in
housing prices than Germany.
As proposed by Schmidt
CB FS
FSec RE
Credit channel
Germany
United Kingdom
Interest rate channel
Germany
United Kingdom
As revised by Goodhart
Credit channel
Germany
United Kingdom
Interest rate channel
Germany
United Kingdom
Figure D6.1 Monetary policy effectiveness in Germany and the United Kingdom
243
244
18
16
14
12
10 United Kingdom
Kingdom
Germany
8
0
1980:1 1982:1 1984:1 1986:1 1988:1 1990:1 1992:1 1994:1 1996:1 1998:1
Figure D6.2 Nominal money market rates, 1980:1±1998:1

10
United Kingdom
2 Kingdom
Germany
–0
–2
–4
–6
1980:1 1982:1 1984:1 1986:1 1980:1 1988:1 1990:1 1994:1 1996:1 1998:1
Figure D6.3 Real money market rates, 1980:1±1998:1
245
246 Charles A. E. Goodhart
Table D6.1 A descriptive statistical analysis of nominal and real money market rates for
the United Kingdom and Germany, l980±l998
Mean Standard Coef®cient

deviation of variation
UK nominal money 9.92 3.33 1.12

market rate
German nominal 6.21 2.43 0.95
money market rate
UK real money 4.09 2.37 1.37
market rate
German real money 3.36 1.24 0.46
market rate
But the truth is that I do not have a good, cast-iron, explanation for my
quandary ± which is, to repeat, that, although I believe monetary policy bites
harder on output in the United Kingdom than in Germany, such policy has
nonetheless been more variable in the United Kingdom.
Let me conclude with a brief comment on dynamics. Because Schmidt sees
the current static position on the effectiveness of monetary policy in the
United Kingdom and Germany as balanced, he is more worried about dynamic
shifts in structure causing future problems. I do not share that worry. I worry
that doubts whether òne size ®ts all' in monetary policy in the Euro Zone
could be initially exacerbated by different ®nancial structures, with adverse
political repercussions. So I view a growing homogenisation of ®nancial
structures across Europe with more and longer ®xed-term mortgages in the
United Kingdom and more reliance on securities markets in Germany as an
almost unalloyed bene®t.
Discussion
Alain Vienney
After a few theoretical reminders concerning monetary policy transmission

channels, Schmidt in Chapter 6 examines the extent to which the structural
differences between the German, British and French systems may affect the
conduct of the single monetary policy.
Schmidt makes the traditional distinction between the German model of
the ®nancing of the economy based on a privileged relationship between
banks and ®rms ± a model in which the vast majority of the ®nancing and
investments of non-®nancial agents are intermediated by the banking system
± and the British market-oriented model ± where the ®nancial markets account
for a bigger share of ®nancing and investment. This contrast is still valid today
whereas the French ®nancial system has to a great extent already moved from
the German to the British model, and the process is probably accelerating.
Favouring an analysis in terms of the `credit channel', Schmidt estimates that
while the share of intermediated ®nancing is larger in the German than in the
British model the elasticity of bank lending to the net situation of borrowers is
signi®cantly lower: German banks favour long-term relationships with
companies and accordingly `smooth out' their lending.
In all, according to Schmidt the German and British models present two
structural differences that cancel each other out, with the result that the
impact of monetary policy on the real economy does not differ signi®cantly
between the two economies. The widening of the Euro Area to include the
United Kingdom is therefore not a problem, and that is good news.
So Chapter 6 makes an interesting contribution to the debate on monetary
policy transmission mechanisms in Europe. In particular I appreciated the
presentation of the ®nancial system types which include the inter-relationship
between sub-systems in a coherent and stable way. It helps to better
understand the monetary transmission mechanism especially when one
considers the credit channel.
However, it seems to me ± and that is one of the major points I would like to
make ± that the analysis shows a somewhat exaggerated bias towards the
`credit channel' and as a consequence, at both the theoretical and empirical
247
248 Alain Vienney
levels, underplays the interest rate channel, which is by far the predominant
one and explains a number of differences between countries. To judge by the
number of stars allocated in Figure 6.4 (p. 228), Schmidt regards the credit
channel as even more important than the interest rate channel, which comes
as a surprise.
1 The interest rate channel ± which is generally believed to

account for most of the differences between the United Kingdom
and continental Europe ± is underestimated
It seems to me that Schmidt pays insuf®cient attention to certain essential

features of the interest rate channel. He does not, for example, mention the
differences in the net position of non-®nancial agents (private individuals in
Germany and France are net creditors ± that is, their assets exceed their
liabilities, while private individuals in the United Kingdom are net debtors1 ),
even though they may explain many of the differences between the United
Kingdom and continental Europe. Thus, in the United Kingdom income
effects and substitution effects have the same impact on consumption (a rise
in interest rates depresses consumption in terms of both the substitution and
the income effect), whereas in the countries of continental Europe the
substitution effect has an opposite impact to the income effect (through the
substitution effect an interest rate rise curbs consumption but increases the net
®nancial income of private individuals and thus their disposable income), and
so the effect of monetary policy on private consumption is less determinate.
Although the author refers to differences in the nature of ®nancial contracts
(®xed rates as opposed to variable or adjustable rates2 ), they are not cited as an
important source of the differences between the United Kingdom and the
countries of continental Europe. It is obviously related to the relative share of
short-term credit where the adjustment of rates can take place more rapidly.3
If you take into account that element of adjustability for the whole balance
sheet of banks, you end up with a net position in terms of adjustable or non-
adjustable rates for assets or liabilities and even if this situation can be dealt
with through appropriate off-balance sheet operations it has a bearing on the
way in which the changes in interest rates are passed onto the clients.
Likewise, the `wealth effects' are not mentioned, although in view of the
different degrees of `marketisation' of the economy, these represent another
major difference between the Anglo±Saxon countries and the countries of
continental Europe. They are clearly evident in the United Kingdom, for
example,4 but are indiscernible in continental Europe according to most
empirical studies.
More generally, the contrast which Schmidt draws between the effect of
monetary policy on the ®nancial system and the effect of changes in the
environment of the ®nancial system on the real economy seems to be of
relatively limited practical relevance. Besides, it is more usual to distinguish
Discussion 249
between the impact of monetary policy on the ®nancial economy (that is the
in¯uence of movements in key rates on bank lending rates and on short- and
long-term interest rates) and the impact of the ®nancial economy on the real
economy.
2 The importance of the credit channel tends to be

overestimated, without looking closely to the formation of bank
lending rates
Although I am not in a position to judge the signi®cance of the credit channel

in Germany, the only evidence of its existence in France is indirect.5 As I have
already indicated, an analysis of the method whereby bank lending rates are
formed (that is their sensitivity to changes in key rates), which does differ
between countries,6 would have been useful. I would like in this regard to
emphasise two elements:
. The net interest margin may allow some leeway on the bank lending rates
and it still differs widely in the Euro Area
. The presence of regulated rates is of primary importance for the setting up
of lending rates, in France, one-third of M3 is made up of this kind of assets
with administrated rates and it is part of the public debate to argue that
these rates prevent an additional lowering of the borrowing terms.
Nonetheless, the analysis of the reasons why German banks, unlike UK

banks, `smooth out' their lending (pp. 225±9) is of obvious value.7 It cannot
be denied therefore that, if we are considering only the credit channel, the
monetary policy transmission channels in the United Kingdom and in
Germany would converge on account of two structural differences that
cancel each other out.8
3 The UK economy is more responsive to monetary policy

decisions than the German or French economies
Simulations conducted, under the aegis of the BIS, by the central banks using
their national macroeconomic models and collated by Smets9 demonstrate a
greater responsiveness to a `standardised' monetary shock in the case of the UK
economy than in the case of France or Germany. Thus, a one-point increase in
the key rate over two years leads in the very ®rst year to a 0.85 per cent
contraction in GDP in the United Kingdom, compared with more limited
declines of 0.4 per cent and 0.37 per cent, respectively, in France and
Germany.10
While the VAR models11 produce more contrasting results, they also
highlight differences in the scale of the adjustment that economies make to a
monetary shock, although the speed of adjustment is more or less the same
250 Alain Vienney
from one country to another (in particular, no price effect is evident within a
period of under eighteen months±two years).
It will be noted, however, that the results for the United Kingdom are very
close to those obtained for Italy and that, in this respect also, there is no
essential difference between the United Kingdom on the one hand and
continental Europe on the other.12
4 The introduction of the Euro will change progressively the

basic structures of the ®nancial systems of the participating
countries
. First, as it is noted in Chapter 6, the overall transmission mechanism is not

that different between the United Kingdom and other European countries.
. Second, monetary policy transmission channels in the United Kingdom
and in continental Europe should gradually converge, in line with the
prospects of the United Kingdom joining the Euro Area and the
development of a large uni®ed capital market in Europe,13 moreover we
do have the example of France moving from a German type to an Anglo±
Saxon type of ®nancial system:
± In many respects the particular features that characterise the channels of

transmission in the United Kingdom re¯ect the high degree of `market-
isation' of the economy (widespread holding of ®nancial assets, which
explains the wealth effect, and lower rate of intermediation than in the
countries of continental Europe). This trend could be followed in
continental Europe as a result of a uni®ed ®nancial market.
± In addition, the United Kingdom integration into the Euro Area could
encourage convergence in interest rate references, which up to now have
been in¯uenced by past in¯ation performance. UK banks might, for
example, increasingly lend at long-term ®xed rates, like their counter-
parts in continental Europe. The latest available statistics would seem to
con®rm that this is, in fact, what is happening.
± Competition will be the driving force of the evolution and it will
progressively generate more similar ®nancing-supply response, means of
payment and asset investments. Finally I dare say that these structural
evolutions will probably be conducive to a certain degree of harmonisa-
tion of the legal and ®scal frameworks in the member countries.
Notes
1. In 1995, the net position in interest-bearing assets were 48 per cent, for German
households, 26 per cent for France and ±9 per cent for the United Kingdom.
2. As a general rule, the United Kingdom prefers variable or adjustable rates (for
example, in the case of mortgage lending), whereas most loans and assets in
Germany and France are still at ®xed rates.
Discussion 251
3. It should be noted that the proportion is the most important in Italy.

4. An increase of GBP 100 in a private individual's net stock market assets leads to a
Great Britain Pound 5 increase in consumption within two years.
5. See for example the article by Barran, Couderc and Mojon (1994).
6. See, for example, Borio and Fritz (1995) and Browne and Fell (1996). The
differences are mainly attributable to the existing level of competition among the
banks and, in France, to the presence of regulated rates. On this subject, see P®ster
Grunspan (1998).
7. Schmidt argues that German banks prefer to take a long-term view of their
relationship with their clients and therefore do not shorten their sails at the ®rst
sign of trouble. This is particularly true since they are often shareholders in the
companies and that, given the long-term nature of the relationship, they are in a
better position than UK banks to assess precisely the effect of a rise in interest rates
on the ®nancial situation and therefore solvency of their clients.
8. The ®rst structural difference is that, in the case of United Kingdom, credit is less
likely to be constrained since ®rms are able to have recourse indifferently either to
markets or to banks.
9. Smets (1995).
10. Simulation conducted with an endogenous exchange rate, except for the bilateral
FRF/DEM rate, which is assumed to be constant.
11. See, for example Gerlach and Smets (1995) or Dornbush, Favero, Giavazzi (1998).
12. The fact that the Italian economy responds so sharply to monetary shocks is largely
due to the importance of short-term ®nance and investment, since the elasticity of
short-term interest rates to key rates is more marked than that of long-term rates.
13. Although Schmidt does not accept this argument (Ì do not believe that the
introduction of the Euro and even the enlargement to include the United Kingdom
will change the basic structures of the ®nancial systems of the participating
countries in any fundamental way' Df. 334, p. 000), he does not explain why.
References
Barrah F., V. Coudert and B. Mojon (1994) `Transmission de la politique mone taire et
credit bancaire, une application a cinq pays del', OCDE CEPII Morning Paper, 94±03,
June.
Borio, C. E. V. and L. S. Fitz (1995) `The Response of Short-term Lending Rates to Policy
Rates: A Cross-country Perspective', Basle, BIS, May.
Browne, F. and L. Fell (1996) `The Changing Institutional Context of Monetary Policy
and its Implication for In¯ation', Frankfurt, EMI, May.
Dornbusch, R., C. Favero and F. Giavazzi (1998) À Red Letter day', CEPR Discussion Paper,
1804.
Gerlach, S. and F. Smets (1995) `The Monetary Transmission Mechanism: Evidence from
the G7 Countries', Basle, BIS.
P®ster, C. and T. Grunspan (1998) `Bank Restructuring, Monetary Policy Ef®ciency and
Financial Stability in France', Central Banks' Economists' Metings, Basle, 28±29
October.
Smets, F. (1995) `Central Bank Macroeconomic Models and the Monetary Policy
Transmission Mechanism', Basle, BIS.
7
European Labour Markets and the

Euro: How Much Flexibility Do We
Really Need?
Michael C. Burda
1 Introduction
In addition to evidence on the nature and source of regional ¯uctuations,

European Monetary Union (EMU) will also provide economists with valuable
new evidence on the monetary transmission mechanism. Given the scepti-
cism with which macroeconomics currently regards monetary policy, current
concern over real effects of EMU comes as a surprise; in a world of ¯exible
prices, space-spanning contingent claims markets and complete information,
it is dif®cult to see why monetary union matters at all for real integration
processes already underway.1 For example, if the real business cycle paradigm
(RBC) ± which emphasises disturbances and propagation mechanisms in the
non-monetary economy and ignores nominal rigidities ± is approximately
correct, the EMU exercise is nothing but a sophisticated veil. To the extent that
EMU leaves ®scal policies and real behavioural incentives unchanged, the
effects of a common currency are of second order at best. In short, this chapter
has no real reason to be written.
Yet, the liveliness of the contemporary debate ± among reasonable and cool-
headed economists for the most part ± is suggestive of an expectation that, for
whatever reasons, real effects of EMU are in the cards. If this is indeed the case,
the underlying presumption must be that nominal disturbances to aggregate
demand and the money supply, in particular, can in¯uence the short-run path
of output and employment, and will continue to do so after EMU is up and
running. Not wanting to make my life too easy, I have decided to write this
chapter from the perspective of an eclectic who is willing to entertain new-
Keynesian arguments. These arguments are important, as the survival of
monetary union will rest on factors outlined long ago by Mundell (1961) and
McKinnon (1963). In Europe, these are perceived to originate chie¯y in labour
markets. From a point of departure that money and monetary policy can
in¯uence real variables, I will discuss the macroeconomic impact of labour
market rigidities on real and nominal adjustment to disturbances in
Èuroland'. However, the most interesting aspects involve taking the
252
European Labour Markets and the Euro 253
discussion one step further: for a number of reasons, the arrival of EMU will
itself have signi®cant effects on the functioning of labour and product markets
and the relative importance of real and nominal rigidities. These feedbacks will
ultimately affect the way Europe reacts as a macroeconomic entity to demand
disturbances and how its central bank views the effectiveness of monetary
policy.
This chapter surveys a number of issues too involved to be treated in model-
theoretic detail here. I will furthermore abstain from econometric analyses for
reasons which should be clear to all. There is a sense that the macroeconomic
regime has changed in a way it has not in several hundred years in Europe: if
the Lucas Critique has any relevance at all, it had better be here and now. I will
adduce some empirical evidence however, which is suggestive of what one
might expect in the future. The chapter is highly speculative, but meant to be
so.
My discussion is organised as follows. In Section 2, I discuss the
macroeconomic impact ± at both regional and pan-European level ± of the
current structure of labour markets. Section 2 surveys the multifarious means
by which a monetary union could affect the functioning of labour markets.
This feedback takes some surprising turns, and may lead to a wholly different
perception of the transmission channels of monetary policy in Europe. Section
3 adduces simple but striking evidence in support of my hypotheses and
Section 4 concludes.
2 How will labour market in¯exibility affect the

macroeconomics of Èuroland'?
Real rigidities and regional ¯uctuations

Robert Mundell taught us long ago that the key to a monetary union's success
can be found in the synchronisation of underlying economic fortunes and,
barring this, the mobility of factors of production, especially that of labour.2
Naturally, labour mobility is costly for both natural and man-made reasons,
and immobility may be regarded differently across cultures and traditions.
Abstracting from social valuations of immobility, losses of output and welfare
are involved when labour does not move to job opportunities, in a geographic,
industrial or occupational sense. To the extent that regional shocks ± such as
an oil discovery in the North Sea or German uni®cation ± continue to occur,
they will wreak macroeconomic havoc on the real evolution of output,
employment and other important variables in ways which are now well
understood. The lack of a ¯exible nominal exchange rate in a world of nominal
rigidities may imply protracted adjustment to regional shocks, unless labour
and other resources move to follow better economic fortunes.
Indeed, the available evidence on labour mobility in the European context is
remarkably discouraging and suggests that a major component of rigidity
254 Michael C. Burda
derives from labour's unwillingness to move.3 In addition, Europe is

characterised by less in-migration, lower fertility and older demographic
structure; all these factors further tend to increase immobility. It would almost
seem unfair to compare Europe with the United States, given that the gene
pool of the latter constitutes a selection of those of the former who had the
strongest incentives to migrate! At the same time, it is worth noting that even
within national boundaries, European labour mobility is low and not capable
of erasing regional disparities, so it is unrealistic to expect much here.4
Yet factor mobility in a monetary union is not restricted to labour, and
under conditions of constant returns one should be indifferent whether the
capital migrates to labour or labour migrates to capital.5 In theory, EMU will
liberate capital mobility as exchange rate risk vanishes, and in fact intra-
European capital mobility has surged in recent years. This is documented in
Table 7.1, which shows the evolution of intra-EU foreign direct investment
(FDI) ¯ows since the 1980. The persistent boom in European equities can be
seen in part as a reaction to the increased mobility now afforded to capital by a
common currency and increasingly integrated asset markets, combined with
ef®ciencies offered by a uni®ed market for goods and services. Whether mobile
capital can smooth out ¯uctuations is not well understood; it stands to reason,
however, that capital should move to places where labour is in excess supply
and could in principle perform this function.
Table 7.1 Intra±EU FDI Flows, 1985±94 (per cent of GDP)
Direct investment in¯ows Balance of direct investment

Country from EU countries to other EU countries
1985±9 1990±4 1985±9 1990±4
Ireland (0.32)a (0.13) n.a. n.a.

Portugal 1.01 1.72 0.96 1.38
Spain 1.02 1.54 0.81 1.24
Sweden 0.26 1.11 ±1.25 ±0.69
Denmark 0.39 1.05 ±0.27 ±0.05
Netherlands 0.91 1.29 ±0.26 ±1.34
Belgium/Luxembourg 1.64 3.05 0.36 0.73
United Kingdom 0.84 0.69 ±0.01 ±0.17
Austria 0.24 0.35 0.07 ±0.08
Italy 0.24 0.19 ±0.03 ±0.17
Greece 0.21 0.53 n.a. n.a.
Finland 0.23 0.47 ±0.73 ±0.75
Germany 0.17 0.11 ±0.28 ±0.62
France 0.42 0.67 ±0.19 ±0.26
Note:
a
Numbers in parentheses are described as highly unreliable.
Source: Dohse and Krieger±Boden (1998).
Product market integration is potentially more important than either form

of factor mobility. Heckscher±Ohlin trade theory under incomplete specialisa-
tion implies that harmonised product prices in traded output produced with
the same technology leads to wage convergence (the factor price equalisation
(FPE) theorem). Consequently the need for factor mobility is eliminated and
the market spreads shocks automatically across the currency area. Here
evidence by von Hagen and Neumann (1994); Fatas (1997); Frankel and Rose
(1996); Bayoumi and Eichengreen (1993, 1996) and others seems to point to
increasing product market integration over time, although this literature has
tended to emphasise quantities more than prices.
Nominal frictions, real rigidities and pan-European macroeconomic

¯uctuations
The next point of discussion is the role of nominal frictions in the European
context. What could the sources of non-neutralities of money in a future EMU
be? Arguing from the status quo, the common perception is that nominal
rigidities play a subordinate role in European business cycles. The standard
assumption is that the large role of centralised collective bargaining, the use of
indexation and a high degree of openness all made Europe more likely to
translate demand disturbances rapidly into price level changes than the
United States, Canada, or Japan. A thorough if somewhat dated discussion of
these issues can be found in the work of Michael Bruno and Jeffrey Sachs,6
who distinguished between US and continental European labour markets by
their reaction to nominal demand and supply shocks. For them, the structure
of labour markets ± meaning to a large extent institutions of wage
determination ± was a key determinant of adjustment to macroeconomic
and especially supply-side disturbances.
As this chapter's title suggests, the functioning of the labour market will be
central to understanding the effects of EMU.7 Mainstream macroeconomics
predicts real effects of money and nominal demand ¯uctuations when
impediments prevent the clearing of product and especially labour markets.
While the origin of these impediments is still poorly understood, it is also clear
that the role of rigidities in nominal and real spheres are highly complemen-
tary for any neoclassical or `new-Keynesian' account of macroeconomic
¯uctuations (e.g. Ball and Romer, 1990; Blanchard, 1990; Romer, 1996;

Jeanne, 1998; Roger, 1998; Kollmann, 1999). This means that it is not
suf®cient for nominal rigidities (such as menu costs) to exist, but they must
also exist alongside real rigidities. In one widely cited mechanism, coordination
failures prevent agents from moving the economy to a better equilibrium.
This complementarity lends intuition to Milton Friedman's (1953)
argument for ¯oating exchange rates. In a famous analogy, Friedman
compared the gains from ¯exible rates to those from setting all clocks back
one hour in the fall and forward in the spring: it is more ef®cient to change the
nominal time standard (the nominal exchange rate) than it is to require
Figure 7.1 Complementarity of real and nominal rigidities for a given price change
millions of individuals to adjust their daily time schedules (nominal domestic

prices) to the annual solar cycle (changing demand and supply conditions).8
Blanchard (1990. pp. 810ff.) and especially Romer (1996. p. 283) make the
reasoning more explicit: individuals do not change their nominal schedules in
the absence of daylight savings time because of the real costs they incur, given
that all others do not change their behaviour.
We are dealing with ®rms which set prices. The extent of real rigidities for a
given price change can be thought of as the resource cost to ®rms of not
moving to optimal pricing in the absence of nominal frictions. In the two panels
of Figure 7.1, this is given by the shaded areas, which are approximately
triangles with base Q � Q* (equal to the output difference between passive
quantity adjustment at rigid price p given by Q and the pro®t maximizing
output level given by Q*), and height equal to the gap between marginal cost
(MC) and marginal revenue (MR) at output level Q. The latter depends on
various factors such as the behaviour of the marginal product of labour,
marginal capacity costs, and the elasticity of labour supply. In the panel (a),
the costs of not changing price from p to p* are relatively small, since the
desired quantity change is modest and marginal costs are ¯at.
In contrast, the ®rm depicted in panel (b) of Figure 7.1 is under considerable
cost pressure to change prices, as can be seen by the vertical difference between
marginal revenue and price for the last units produced. Passive quantity
adjustment implies a large departure from unconstrained optimal production
Q*, while sharply rising marginal costs means that these additional units are
being produced at a large loss.9 For a given costly nominal price adjustment,
the ®rm in panel (a) is likely to maintain rigid nominal pricing, while the ®rm
in panel (b) will adjust its prices. Comparing the two panels, one sees the
necessity of real rigidities: individual ®rms have little incentive to change
prices, given that others are not doing so. Strategic complementarity implies
that second-order issues for the ®rm can have ®rst-order effects for the
macroeconomy.
Money wage rigidity could also induce business cycle ¯uctuations. While an
important element in the early intellectual development of Keynesian
macroeconomics, nominal wage rigidity is not borne out at the micro level
(Bils, 1985, Smith, 1999) nor is it particularly supported by aggregate evidence
on wage and price dynamics (see references in Blanchard, 1990); Jeanne (1998)

and Roger (1998) have both shown that nominal price rigidity, combined with
some degree of real wage rigidity, is suf®cient to generate persistent
¯uctuations that resemble US business cycles.10
Summary
The previous discussion can lead to rather sombre conclusions about the
future of EMU. First, the conventional wisdom of extreme rigidity in labour
markets, which now has the OECD seal of approval (OECD, 1994) and is
accepted nowadays by everyone except the labour unions and perhaps a few
surviving extremists in the German ®nance ministry, should render the EMU a
Mundellian nightmare. It won't be necessary, according to this logic, for
another German reuni®cation to occur to generate real problems. All we need
is some overheating in Ireland, Portugal, or Finland and the whole EMU
project will collapse as the other regions slump without any equilibrium
mechanism.
An equally pessimistic message emerges on the monetary transmission
mechanism when considered under these circumstances, in which a rapid
pass-through into in¯ation is taken for granted by market participants.
Reviews by Buti and Sapir (1998) and Dohse and Krieger-Boden (1998) give
rather sombre pictures of the prospects, and Dornbusch, Favero and Giavazzi
(1998) raise questions about the asymmetric impact of monetary policy on the
participating EMU countries. Moreover, ®scal policy is hamstrung by the
Maastricht Treaty and the Pact for Stability and Employment and potential
exists for `beggar-thy-neighbour' effects as countries jockey to better their
macroeconomic circumstances. This `Flassbeck±Lafontaine Hypothesis' sees
purposeful competitive de¯ation just around the corner, as countries unable to
devalue are forced to regain competitiveness by more painful means. In this
view, governments, robbed of their power to generate instant nominal
devaluations will do what Britain did in the ®rst half of the 1920s. Feldstein's
(1992) criticism is now widely accepted that politics have outweighed
economics; Eichengreen (1998) has already speculated about the `dissolution'
of the European Monetary Union before it even begins.
Given this doomsday scenario, critical economists are compelled to ask the
question: Are rigidities in Europe set in stone? Is it reasonable to assume that
the Euro will leave labour markets and their institutions intact ± and, if not,
which ones are implicated? What will be the consequences of these changes?
What follows is a highly speculative discussion of three areas: (1) nominal

rigidities; (2) real rigidities, holding institutions constant; and (3) changing
institutions.
3 Will the Euro affect labour market ¯exibility?
Nominal price rigidity should increase

First, I speculate that a number of factors will cause nominal rigidities to
increase in Euroland, especially that of nominal prices. First, the introduction
of a common currency will effectively convert a Europe of many small open
economies into a behemoth with an import ± export exposure of 10 per cent of
GDP, roughly as closed as the United States and Japan. This is a regime change
of striking character. As a consequence, a large share of industry will be moved
into the `home goods' sector, and will no longer be exposed to vagaries of
nominal exchange rate and international demand ¯uctuations. For small,
open economies with output more likely to be concentrated in the value-
added chain, exchange rate disturbances are re¯ected rapidly in both input
and output prices; a monetary union in Euroland removes this aspect, as
inputs become increasingly non-traded goods invoiced in Euros. Devaluation-
induced expenditure-switching is no longer possible on a grand scale.11
Factors favouring nominal rigidities ± i.e. customer relationships, search costs,
etc. ± should become relatively more important than costs associated with
cross-border transactions.12 Cost pressures will increasingly be restricted to
domestic (Euroland) labour markets, marginalising the importance of
exchange rate changes for pricing decisions.13 Figure 7.2 illustrates how the
reaction of local currency costs to a devaluation are decisive in determining
incentives to adjust prices. In panel (a), which corresponds to a small open
economy, marginal costs rise in response to a devaluation and the incentive to
change prices rise commensurately. In panel (b) ± which corresponds to
Euroland ± the incentive is less strong, leading to a larger output effect.
A second effect is more subtle (and possibly less relevant). A common
currency area is generally assumed to increase competition, as improved price
transparency opens up national markets to intra-EMU, cross-border rivals. At
the same time, however, monetary union in Europe necessarily implies a
signi®cant decrease in the overall relevance of the external market for the
representative producer. Assuming that foreign trade is perfectly competitive
and priced off the exchange rate according to the law of one price, the
representative exporting ®rm pre-EMU, ironically, may face an enlarged
domestic market with more pricing power on balance, to the extent that the
market using Euros increases relative to that using foreign currencies. This is
especially true if the pace of mergers and acquisitions within Euroland
continues. To the extent that ìnwardisation' increases monopolistic power in
price-setting, it will increase incentives not to adjust prices in the domestic
currency, for reasons stressed by Akerlof and Yellen (1985), Mankiw (1985)
and Romer (1996). Increased exposure to the sheltered domestic market will
mean greater incentives to price to market and to set nominal prices in
advance for longer periods, as customer relations become more important and
the net bene®ts of charging stable nominal prices increase (Okun, 1981).
The third and potentially most important effect ¯ows from the credibility
that comes from having a central bank which can `stand above' (i.e. ignore)
economic conditions in individual countries and be free of political pressure.
To this extent if the European Central Bank (ECB) is really the most
independent central bank in the world, agents will be more prone to expect
low in¯ation and will not attribute deviations to policy changes. This
important source of inertia should be distinguished from the usual wage±
price mechanism (e.g. Blanchard, 1990); rather it has to do with the anchoring
of in¯ationary expectations and the effect this will have on the willingness to
negotiate contracts in nominal terms.
To give some sense on the evolution of rigidities, I present some simple
statistics for data on comparable price and wage time series from EU member
countries.14 Table 7.2 displays average unweighted correlations of bilateral
in¯ation rates (®rst difference in the logarithms) for a number of groupings of
countries in addition to the Euro-11 since 1961. For comparison, I present data
for eight regions of the United States for a similar time period. Clearly, an
increase in price convergence has taken place across the board, not only in the
smaller `core' groupings. The eigenvalues of the moment matrix indicates the
extent to which in¯ation in one country can be expressed as a linear
combination in others. Table 7.3 documents that, to a large extent, my
Figure 7.2 The cost of passive quantity adjustment in response to an exchange rate
depreciation
Table 7.2 Synchronisation of price in¯ationa in Europe and the United States, 1961±96
Average correlation coef®cient in Smallest and largest moment matrix

group (std. dev.) eigenvalues (1961±79) and (1980±96)
Total 1961±79 1980±96 1961±79 1980±96 Percentage

sample change
Core Europe 0.76 0.80 0.82 9.82 10�4 5.89 10�4 ±39.9
(B, NL, D, A) (0.08) (0.06) (0.09) 0.207 0.0928 ±55.2
Core Europe 0.74 0.71 0.81 8.74 10�4 1.64 10�4 ±81.2
F, DK, IT (0.11) (0.13) (0.12) 0.560 0.363 ±35.1
Euro-11 lightb 0.73 0.73 0.80 6.53 10�4 3.88 10�5 ±94.1
(0.14) (0.14) (0.15) 0.983 0.602 ±38.8
Euro-11 lightb 0.71 0.70 0.78 3.17 10�4 3.69 10�5 ±88.4
DK, S, UK (0.13) (0.15) (0.14) 1.336 0.811 ±39.3
Memo: USA
8 regions, 0.95 _ _ 1.48 10�5 _ _
1978±92, (0.03) 0.376
GSP de¯ator
Notes: a In¯ation is measured as ®rst difference in the logarithm of the relevant price index.
b
Less Luxembourg, Portugal.
B Belgium; NL Netherlands; D Germany; A Austria; F France; DN Denmark; IT Italy;
S Spain; UK United Kingdom; USA United States.
Sources: US Bureau of Economic Analysis (REIS), IMF, International Monetary Statistics.
conclusions hold when looking at a much smaller time interval and when
correcting for exchange rate changes.
It has been argued, by Calmfors (1998a) and others, that monetary union
could result in increasing nominal money wage rigidity. Presumably this
would arise as a result of the low level of in¯ation and resistance to nominal
wage reductions. In addition, the alignment of traded goods prices should
impose factor price convergence, as long as complete specialisation does not
occur ®rst, although this can only be a statement about labour of a given
quality. At the same time, Calmfors (1998a) claims that increasingly variable
macroeconomic conditions might lead to shorter nominal contract periods
and greater nominal wage ¯exibility.
Nominal wage behaviour in Europe over the past thirty years lends support
to my contention that nominal wages are less likely to be rigid than prices.
Table 7.4 and 7.5 clearly show a determination in the strong positive
correlation of real wage growth present in the 1960s and 1970s. To the extent
Table 7.3 In¯ation correlations, in national currency and DM terms, 1976±96
Average correlation coef®cient in group
Annual OECD in¯ation rate Annual OECD in¯ation rate in DMa

terms using BLS exchange rates
Total 1976±86 1987±96 Total 1976±86 1987±96

Sample Sample
Core Europe 0.82 0.81 0.77 0.56 0.52 0.70

(B, NL, L, D, A) (0.10) (0.08) (0.12) (0.27) (0.31) (0.17)
Core Europe 0.80 0.79 0.33 0.45 0.45 0.38

F, DK, IT (0.11) (0.12) (0.50) (0.23) (0.25) (0.41)
Euro-11 0.79 0.67 0.48 0.44 0.48 0.37

(0.13) (0.26) (0.37) (0.21) (0.24) (0.40)
Euro-11 0.78 0.67 0.45 0.49 0.54 0.38

DK, S, UK (0.12) (0.24) (0.41) (0.21) (0.22) (0.42)
Notes: a OECD in¯ation corrected using BLS exchange rates.
For country codes, see Table 7.1. Standard deviations in parentheses.
Source: OECD.
that increasing èntropy' in the behaviour of nominal wage movements is

re¯ected by decreasing cross-country correlation, this supports the assertion
that nominal wage ¯exibility is increasing, not decreasing over time. The
largest eigenvalue of the moment matrix for ®rst differences in nominal wages,
compared with that of nominal prices, is larger and the decline in the
eigenvalues are smaller, suggesting that nominal wages in this context do not
seem to merit the description `rigid'.
Not only are nominal wages less correlated across European countries
than US regions but their levels have exhibited divergence in the past
decade. Table 7.6 displays US BLS data on hourly compensation in the EU
and computes coef®cients of variation for the groupings CORE (Belgium,
Luxembourg, France, Germany and Austria); CORE Denmark France
Italy; The EURO-11 (CORE plus Ireland, Finland, Spain, Portugal, France and
Italy). For each grouping Germany was retained and dropped to examine the
in¯uence of that country, especially in light of German uni®cation. In all
cases except the CORE less Germany (the Benelux countries plus Austria),
the cross-country variability of nominal wages increased over the ten-year
period.
Table 7.4 Synchronisation of nominal wage growtha in Europe and United States,
1961±96


sample change
Core Europe 0.85 0.76 0.46 2.15 10�3 1.24 10�3 ±42.3
(B, NL, D, A) (0.06) (0.17) (0.11) 0.688 0.130 -81.0
Core Europe 0.72 0.52 0.48 1.49 10�3 6.59 10�4 ±55.8
F, DK, IT (0.15) (0.32) (0.18) 1.49 0.451 ±69.8
Euro-11 lightb 0.71 0.46 0.55 5.58 10�4 2.57 10�4 ±54.0
(0.15) (0.35) (0.22) 2.39 0.760 ±68.2
Euro-11 lightb 0.66 0.48 0.50 1.80 10�4 6.07 10�5 ±66.2
DK, S, UK (0.18) (0.31) (0.26) 2.96 0.981 ±66.9
Memo: USA
8 regions, 0.92 _ _ 2.01 10�5 _ _
1978±92, (0.06) 0.449
annual comp.
Memo: USA
8 regions, 0.90 _ _ 1.65 10�5 _ _
1978±92, (0.08) 0.425
wages/salaries
Notes: a Nominal wage growth is measured as ®rst difference in the logarithm of the wage index.
b
less Luxembourg, Portugal.
For country codes, see Table 7.1.
Sources: US: Bureau of Economic Analysis (REIS), IMF, International Monetary Statistics.
Real rigidities should decrease given current institutions

It is interesting that there are so many who believe that real rigidities in Europe
threaten the success of monetary union, and I am sure that my invitation to
contribute to this volume was related to my perceived views on real rigidities
in European labour markets. Indeed, a number of arguments can be found to
buttress the claim that in¯exibility in the labour market will spell the death of
EMU. Yet how robust are these arguments to the Lucas Critique ± i.e. the
introduction of the Euro? In my view, the more important and subtle effect of
EMU has largely escaped scrutiny: How will a common currency affect the
functioning of labour markets? Could the vaunted lack of labour market
¯exibility in continental Europe be affected by the introduction of a common
currency? If so, how?
Table 7.5 Nominal manufacturing wage growth correlations in national currency and
DM termsa
Annual nominal wage growth in Annual nominal wage growth in DM

manufacturing in local currency basis
Total 1976±86 1987±96 Total 1976±86 1987±96

sample sample
Core Europe 0.68 0.64 0.42 0.44 0.44 0.39

(B, NL, L, D, A) (0.11) (0.12) (0.34) (0.27) (0.26) (0.33)
Core Europe 0.66 0.59 0.22 0.29 0.25 0.24

F, DK, IT (0.12) (0.22) (0.33) (0.25) (0.31) (0.25)
Euro-11 0.68 0.56 0.34 0.30 0.28 0.28

(0.13) (0.19) (0.33) (0.27) (0.32) (0.39)
Euro-11 0.65 0.55 0.30 0.32 0.33 0.26

DK, S, UK (0.13) (0.19) (0.35) (0.27) (0.30) (0.40)
Notes: a First differences in log hourly nominal compensation costs for production workers in
manufacturing, in local currency or in DM converted using annual average exchange rates.
For country codes, see Table 7.1. Standard deviations in parentheses.
Sources: US Bureau of Labor Statistics; author's calculations.
Table 7.6 Nominal wages in manufacturing in the EU, 1986 and 1996
Money wages in Europe in dollars Unweighted coef®cients

(nominal hourly compensation) of variation of nominal wages
Land 1986 1996 Grouping 1986 1996
Luxembourg 10.86 22.55 CORE 0.095 0.143

Belgium 12.43 25.89 (A, B, D, L, NL)
Germany 13.43 31.87
Netherlands 12.22 23.14 CORE less D 0.077 0.064
Austria 10.73 24.95
France 10.28 21.19
Denmark 11.07 24.24 CORE DK, I, F 0.098 0.173
Italy 10.47 17.48
Finland 10.71 24.95
Ireland 8.02 13.85 CORE less D 0.076 0.123
Portugal 2.08 5.58
Spain 6.25 13.40
Sweden 12.43 24.56 EURO-11 0.331 0.358
United Kingdom 7.66 14.13
memo: USA 13.26 17.70 EURO-11 less D 0.336 0.342
Note: For country codes, see Table 7.1; L Luxembourg, I Ireland.
Sources: US Bureau of Labor Statistics, Of®ce of Technology and Productivity.
Because the quanti®cation of real rigidities is dif®cult and undoubtedly

subject to regime changes (Calmfors 1998a) it seemed unwise to estimate
measures of nominal and real wage rigidity, on the other hand it is reasonable
to conclude that for the most part the two pressure points on which all real
rigidities rest are (1) collective bargaining and unions and (2) the social safety
net and especially unemployment bene®ts. My discussion below will
concentrate primarily on these.
The elasticity of labour demand will increase

The ®rst Euro assault on real rigidities is the weakening of union power in wage
determination. While unions are already in retreat in much of the OECD
(OECD, 1994), in Europe this decline is largely restricted to Britain, member-
ship losses in France and Italy belie an ever-strong in¯uence on central wage-
setting institutions, in Germany, membership has declined primarily in the
East, where it was arti®cially high to begin with. Yet the brave new world of
Euroland portends ill for continental collective bargaining, which has always
been a national institution with national idiosyncracies. A simple textbook
argument ± namely, the Marshall±Hicks rule of labour demand ± predicts that
the melding of European nations into a currency union will attenuate unions'
ability to monopolise the supply of labour by increasing the demand elasticity
they face.15
Three of the four elements of the Marshall±Hicks rule will be operative. First,
labour unions derive their attractiveness from their ability to tap into quasi-
rents that their employers can earn in the market. In a globalising Europe,
product market competition among companies operating with quasi-rents will
increase dramatically, which translates into an increase in the elasticity of
product demand and the elasticity of the derived demand for labour.16 Second,
the acceleration of intra-European corporate mergers and takeovers opens up
the possibility of easy substitution of capital and cheaper labour for more
expensive labour within the Euroland area. This attenuates the bargaining
strength of national unions. Third, for any given national labour market, the
rest of Euroland is large (and possibly getting larger), meaning that the supply
elasticities of competing factors is likely to be high.
How will European labour unions cope with these powerful winds of
change? Already hamstrung by fragmentation along industrial, regional, or
religious lines, they will face language and national cultures as further barriers
to their effectiveness. Despite considerable rhetoric, searches of labour union
literature (including the Internet) have yielded little concrete evidence of an
effective Pan-European labour movement. While a similar argument applies to
employer associations, the growing transnationality of capital puts labour at a
clear bargaining disadvantage ± a forced decentralisation in the Calmfors±
Drif®ll (1988) sense. The potential for coordinated bargaining strategies is
presently incompatible with union structures across countries, which is
essential to purposeful pattern bargaining. These structures represent decades
of gradual evolution and cannot be changed overnight. To me, at least, it

seems highly implausible that Europeans will accept wage leadership of
German engineering and public sector workers after having ®nally shaken
themselves from the yoke of Teutonic monetary policy!
Strategic interaction of unions with the central bank will change

The argument that labour market rigidities might be endogenous has been
made by a number of analysts (Danthine and Hunt, 1994, Berthold and Fehn,
1997, Dohse and Krieger-Boden, 1998 among others). While I take the
position that competition will impose decentralisation and deregulation of
EMU labour markets, a number of analyses emphasise changing strategic
interactions between central banks, unions and governments and the effect
these can have on aggregate outcomes. In particular the incentives for unions
to recognise the effects of their wage demands on the macroeconomy stands at
the centre of this discussion.17 An important strand of the literature which has
emerged in the run-up to EMU takes Calmfors and Drif®ll's (1986)
contribution as a starting point, which relates the centralisation of collective
bargaining to the degree to which unions internalise the effect of collective
bargaining on the macroeconomy. Early on, the risks of simply extending this
analysis to the EMU context were made clear by Danthine and Hunt (1994).
They showed that product market integration will play an important role in
¯attening out the `hump' therefore rendering centralisation of collective
bargaining less relevant. Another strand has been explored by Cuikierman and
Lippi (1998, 1999) who look at strategic interactions of the centralisation of
nominal wage-setting and central bank independence.
While these analyses are intellectually stimulating, I am convinced that the
most pressing effects of monetary union derive from the fact that existing
market imperfections and distortions will be subject to forces of competition;
these effects are likely to swamp Barro±Gordon (1993) and Calmfors±Drif®ll
(1988) effects and issues of time-consistency, reputation and coordination. I
would therefore go even farther than Danthine and Hunt (1994) and argue
that structural change implied for labour and product markets needs to be
studied carefully before venturing guesses on the future strategy spaces of
policy makers. It is, of course, the Lucas Critique again: the elasticity of labour
demand will change, the objectives of labour unions will change, their
constraint sets will change, the analyses cited above generally assume
complete product market integration and ignore capital as a competing factor
of production. Local national unions which insist on aggressive wage
settlements will be faced with higher local joblessness. Only if the social
safety net accommodates higher unemployment will unions be able to ignore
these factors, and given the hard budget constraint of the monetary union,
they will ®nd it increasingly dif®cult to do so.
The Euro and labour market institutions

An equally interesting hypothesis is that European jurisdictions will adapt and
possibly reform labour market regulation in light of the increasing pressures
brought about by EMU as well as globalisation and technological innovation.
In this view, increased competition among member EU states as well as among
regions within EU states will lead to a Nash equilibrium in which each member
state disregards the effects its behaviour has on the others. This type of
competition might emerge directly, in which some initiate direct labour
market reforms in the hope of `beating the competition' and reap short to
medium-term gains, the success stories of the Netherlands and Denmark
might be viewed in this light. Another channel is increased tax competition ±
especially, but not only, corporate taxation ± to enhance the attractiveness of
investment in local economies (Standortwettbewerb in the local jargon), as
Ireland has done aggressively in recent years. This tax competition puts strain
on national member country ®nances and may force spending cuts and
structural reforms. The experience of US states in this regard indicate that this
mechanism can be powerful indeed. Bean (1998) has discussed this aspect.18
At the same time, it seems unlikely that the EU Commission and Parliament
will sit idly and watch this `race to the bottom'. Already minimum capital
taxation has been all but agreed to, while the probability of increased
international (intra-European) competition along the social dimension is
severely hampered by the Social Charter, which was rati®ed at the Strasbourg
Summit in 1989 by all EU governments except the United Kingdom.19 The
about-face on fast-track membership of the new market economies of Central
and Eastern Europe may re¯ect a fear that unbridled competition in both
regulatory and tax dimensions might be triggered by early admission of these
countries. Yet the lack of consensus for a federal European ®scal policy means
that little substantive support for harmonisation will come from the top.
Summary
What are the macroeconomic implications of increasing nominal rigidity and
real ¯exibility, ceteris paribus? The empirical evidence, which is meant to be
suggestive, supports the contention that nominal price rigidity has increased
as a consequence of product market integration and exchange rate stability.
Nominal wages, in contrast, are highly correlated only in the core, and this
applies a fortiori to real wages and real exchange rates as well. These ®ndings
suggest that the Euro will affect labour market ¯exibility in the direction of
more ef®ciency. Without more detailed information on preferences, it is
impossible to say whether this increase in ef®ciency will lead to overall welfare
gains; some analyses, such as Agell (1998), claim that labour market rigidities
may re¯ect welfare-improving policies in the light of other market imperfec-
tions. Burda (1995) has presented a related rationale for union wage
compression.
Wage-setting will become more fragmented unless pan-European efforts

arise to coordinate. On the collective bargaining front, managing this change
will require Herculean efforts on the part of national labour movements. In
this vein one could expect a restructuring of unions in France, Spain and Italy
(and possibly the United Kingdom) towards centralised industrial unions in
order to facilitate cross-border cooperation; Dohse and Krieger-Boden (1998)
describe the emergence of Èuropean Works Councils' in large enterprises. Yet
the reality of labour relations in these countries as well as the divergence of the
interests of labour at the national level portend less dramatic changes (Streeck,
1998). While the possibility of pattern bargaining by large industrial unions ±
as in Austria, Germany, or Sweden ± is frequently discussed, it is dif®cult to see
how it could lead to truly coordinated outcomes without a strong central
organisation as is the case in these countries. Because I see pan-European
coordination coming in a decade's time at the earliest, a more modest goal for
organised labour would simply be to get control over the process. The example
of Eastern Germany can be seen as a lesson on how not to do it.
Increasing nominal wage ¯exibility combined with nominal price rigidity is
likely to lead to increased real wage ¯exibility. Casual evidence I have
assembled in Tables 7.7 and 7.8 show that real wage behaviour in EU members
has become increasingly uncorrelated over time, and that this tendency
increases with the size of the group considered. This can be contrasted with US
evidence, which shows a remarkably high correlation given the size of the
regions considered.
The empirical evidence suggests that while there is enough ìnsurance
potential' in many respects to reduce Europe-wide risks, it is not showing up in
wage growth rates. The dramatic deterioration of real wage correlations is
evidence, to my mind at least, that there is potential for ¯exibility, at least
between the core and the rest of the Euro-11. This ¯exibility supports my
contention of a `forced decentralisation', which would not have been less
likely had a two-track solution to the monetary union question been
implemented. The breakdown of the synchronous behaviour of real wage
growth in Germany and Holland in the early 1990s depicted in Figure 7.3 is
one example of how this has occurred.
The macroeconomic implications of increasing nominal rigidity and declining
real rigidity, ceteris paribus, are somewhat surprising. The old conventional
wisdom (Sachs, 1979, 1983, Bruno and Sachs, 1985) was that the United States
was characterised by nominal rigidity but real wage ¯exibility, the nations of
Europe in contrast had real rigidities but not nominal ones, which led to
accentuated responsiveness of nominal wages to aggregate demand movements
and to an attenuation of policy makers' ability to use monetary policy even to
the limited extent now allowed in mainstream macroeconomics. The implica-
tions of my analysis is that Europe is likely to develop a more pronounced,
common cycle as its own response to monetary policy evolves. This is the
conclusion reached by more recent analyses such as Jeanne's (1998).
Table 7.7 Synchronisation of real wage growtha in Europe and the United States,
1961±96


sample change
Core Europe 0.60 0.69 0.24 2.69 10�3 9.04 10�4 ±66.4
(B, NL, D, A) (0.08) (0.16) (0.38) 0.170 0.014 ±91.5
Core Europe 0.59 0.45 0.08 1.76 10�3 2.76 10�4 ±84.3
F, DK, IT (0.13) (0.24) (0.41) 0.291 0.018 ±93.6
Euro-11 lightb 0.55 0.36 0.06 9.96 10�4 1.937 10�4 ±80.6
(0.13) (0.25) (0.42) 0.405 0.026 ±93.5
Euro-11 lightb 0.46 0.35 0.14 5.62 10�4 1.35 10�5 ±97.6
DK, S, UK (0.20) (0.24) (0.39) 0.455 0.036 ±92.1
Memo: USA
8 regions, 0.59 _ _ 6.68 10�5 _ _
1978±92, (0.18) 0.016
real comp.)
US (8 regions,
1978±92, 0.55 _ _ 6.10 10�5 _ _
real wages and (0.20) 0.016
salaries)
Notes: a Real wage growth is measured as ®rst difference in the logarithm of the nominal wage index
reported by the IMF, International Finance Statistics, divided by the IMF/IFS consumer price index.
b
Less Luxembourg, Portugal.
For country codes, see Table 7.1.
Concluding remarks
In addition to its historic dimensions, EMU will shed new light on a number of
old, bothersome questions. Naturally, it will help us understand better how
monetary unions function. In the ®rst instance, however, it will teach
economists and policy makers the relevance of the new-Keynesian approach
to understanding aggregate ¯uctuations, for which there is precious little
evidence. It will also help us decide whether nominal price or wage rigidities
are more relevant for explaining the real effects of aggregate demand
¯uctuations and thus the transmission mechanism itself.
The convergence of exchange rate and especially price dynamics suggests
that the preconditions for nominal price rigidities have become more
Table 7.8 Manufacturing real wage growth correlations using different price indexes
Wages de¯ated by OECD price index Wages de¯ated by IMF price indexa
Total 1976±86 1987±96 Total 1976±86 1987±96

sample sample
Core Europe 0.39 0.49 0.06 0.44 0.50 0.17

(B, NL, L, D, A) (0.25) (0.23) (0.49) (0.26) (0.23) (0.59)
Core Europe 0.22 0.27 0.13 0.23 0.27 0.13

F, DK, IT (0.26) (0.30) (0.38) (0.43) (0.27) (0.43)
Euro-11 0.13 0.14 0.13 0.12 0.12 0.13

(0.25) (0.30) (0.36) (0.25) (0.29) (0.38)
Euro-11 0.14 0.17 0.13 0.14 0.16 0.14

DK, S, UK (0.23) (0.29) (0.35) (0.38) (0.29) (0.36)
Notes: a Luxembourg excluded. Standard deviations in parentheses.
Per cent
14
United States
12
Germany
10 Netherlands
Changes to previous year
–2 1961 1966 1971 1976 1981 1986 1991 1996
–4
–6
Figure 7.3 Growth rates of real hourly compensation costs in manufacturing, 1960±96
favourable in Euroland. At the same time, trends in money±and, especially,

real±wages point to declining importance of real rigidities. Real economic
conditions and institutions are increasingly unfavourable for `business as
usual' in the European union; the breakaway behaviour of the Netherlands,
Denmark and possibly Ireland and Portugal supports the hypothesis that EMU
is a Trojan horse of decentralisation ± not only de facto, but more importantly

for structural reasons related to product and capital market integration.
As many have recognised, the functioning of labour markets is central to the
macroeconomic future of Euroland, but the mechanisms are remarkably
subtle. The most important of my messages can be summarised as follows.
First, the introduction of common currency, price transparency and internal
trade integration ± will lead to a ìnwardisation' of the European continent
with the implication that internal and external nominal shocks will have less
impact on nominal wage- and price-setting, and show up more strongly in
output variation. Second, the standard analysis suggests that this will be
related to the extent the underlying real economy is responsible for output
¯uctuations. In the past continental European countries were known for their
`real rigidities' and in¯ation appeared to respond rapidly to changes in
nominal demand.20
Yet my prediction that the EMU amounts to a `forced decentralisation
programme' which will subject these rigidities to increasing pressure is
accompanied by an optimism that a reduction of these rigidities will follow.
Most important of the forces are increasing capital mobility, trade integration
and competition, which will force wages for labour of given quality to converge
(factor price equalisation) as well as to react more ¯exibly to changing local real
conditions. Labour mobility, while a central point of discussion, is a side show
which is not as relevant in the short run for the United States as it is made out to
be.21 As more continental European countries scale back safety nets, it will
become increasingly dif®cult for real wage determination to stay out in front of
nominal developments. This ¯attening of the Phillips and aggregate supply
curves will facilitate a more potent monetary policy. My prediction is that,
unless an improbable miracle occurs in pan-European collective bargaining,
labour markets will become more and not less ¯exible in the future. Calls for
additional ¯exibility may be the economic equivalent of whipping a dead horse,
and could provoke counterproductive reactions.
As if it were not controversial enough to sell the Euro as the Trojan horse
which liberalised labour markets, I ®nd it highly likely that it will change the
macroeconomics of Europe fundamentally over the next decade and thus
foster in a new regime for ®scal and monetary policy. Monetary policy should
gain a new potency, as Europe begins to look more like the United States and
Japan and less like Germany and France. A new role for monetary policy
should emerge, although the usual caveat remains that the effectiveness of
monetary policy is largely an artifact of its not being used in a predictable way
to in¯ate the economy (Taylor, 1980). Therefore my chapter should not be
construed as endorsing Oskar Lafontaine's `domestic demand strategy', but
rather a warning that the temptation to employ such a strategy will increase in
future years.
Of course, my analysis is predicated on the view that nominal rigidities,
especially price rigidities, are important in the evolution of a macroeconomy
in the short run. If I turn out to be wrong and have to eat my hat, this fact will
nevertheless have been useful information for our profession as well as policy
makers. If I am right, EMU will have delivered the ultimate bonus in real
ef®ciency gains for the unemployment-riddled labour markets of the
continent.
Appendix
In Tables 7.2±7.8, I present some suggestive evidence in support my twin

hypotheses of increasing nominal rigidities on the one hand and decreasing
real rigidities on the other. The variables considered are (1) consumer prices,
(2) nominal wages for the total economy, (3) real wages, all from IMF,
International Financial Statistics and using a longer sample (1961±96), data
gathered by the US Bureau of Labor Statistics <http://stats/bls/gov/progho-
me.htm> on manufacturing wages and exchange rates, and the OECD price
index (1976±96). Correlations of ®rst differences in logarithms of these
variables were examined in different grouping: a core group (Germany,
Luxembourg, Belgium, Holland, and Austria); the core plus France, Italy and
Denmark; the Euro-11; and ®nally the Euro-11 adding back Denmark, plus
Sweden and the United Kingdom. The average correlation coef®cient provides
a rough indicator of co-movement, while eigenvalues of the moment matrix
indicate the extent to which linear combinations of countries' experiences can
replicate others, the number of eigenvalues close to zero later indicates the
extent to which ìnsurance' is possible.
Notes
1. The view that short-run adjustment costs associated with EMU are small relative to
long-run gains has been echoed by Buiter (1995).
2. See Mundell (1961), as well as McKinnon (1963); Kenen (1969).
3. Indirect estimates of labour mobility for the United States by Blanchard and Katz
(1992) and for Europe by Decressin and Fatas (1995) show that European regions
tend to adjust to adverse employment shocks via changes in labour force
participation as opposed to residence. For more detailed summaries of the
evidence, see Eichengreen (1993) and Gros and Hefeker (1998) as well as Obstfeld
and Peri (1998).
4. See Gros and Hefeker (1998) for an overview.
5. It is remarkable that the optimal currency literature has largely ignored the role of
capital mobility ± meaning long run mobility of the means of production ± despite
Mundell's own explicit reference to it in his seminal article. For examples, see
discussions in Bo®nger (1994); Bayoumi and Eichengreen (1996); Wyplosz (1997);
Gros and Hefeker (1998).
6. See Bruno and Sachs (1985); Sachs (1979, 1983); but also Branson and Rotemberg
(1980).
7. Among others, Romer (1996) has stressed the labour market as a primary source of
real rigidities in the macroeconomy, as complementary to nominal rigidities.
8. `The argument for ¯exible exchange rates is, strange to say, very nearly identical
with the argument for daylight savings time. Isn't it absurd to change the clock in
summer when exactly the same result could be achieved by having each individual
change his habits? All that is required is that everyone decide to come to his of®ce
an hour earlier, have lunch an hour earlier, etc. But obviously it is much simpler to
change the clock that guides all than to have each individual separately change his
pattern of reaction to the clock, even though all want to do so.' (Friedman (1953),
p. 173, my emphasis).
9. In fact, the ®rm in panel (b) is more likely to ration output, producing only to the
point at which price equals marginal cost, and thereby violating the assumption of
completely passive (i.e. demand-determined) adjustment of production to demand.
In any case the point is made that incentives to change prices in this case are large.
10. For evidence on the rigidity of prices in the United States see Carleton (1986),
summaries of empirical evidence are available in Blanchard (1990) and Romer
(1996).
11. This argument can also be found in McKinnon (1963), who stresses the role of non-
traded goods in the reaction to devaluations.
12. The failure of ®rms selling into the United States fully to pass through exchange rate
¯uctuations is well documented (see Knetter, 1989; Dornbusch, 1987, 1996) and
could be seen as an indication of what Euroland can expect.
13. One exception could be energy prices, which continue to be denominated in
dollars. As Europe is the largest customer of the oil-exporting Middle East and
Russia it may come to pass that oil prices are denominated in Euros. The relevant
issue, of course, is whether oil prices in Euros will tend to become more stable over
time.
14. The empirical evidence I present in this chapter is rather modest, as it seems foolish to
place much weight on estimates of structures in place before monetary union. On the
other hand many investigators have looked at the temporal evolution of cross-
correlation of price and quantity variables. (For example DeGrauwe, 1991; von Hagen
and Neumann, 1994; Bayoumi and Eichengreen, 1996; Frankel and Rose, 1996). For
details on the data used, the reader is referred to the Appendix (p. 271).
15. The Marshall±Hicks rule states that the elasticity of demand for labour is higher, the
higher the elasticity of demand for output produced with that labour, the higher
the elasticity of substitution between labour and other inputs the lower the
elasticity of supply for those competing inputs and the greater the cost share of
labour in production. See Hamermesh (1993).
16. For evidence on how product market competition has affected labour unions and
labour markets in the United States in general, see Duca (1998).
17. Calmfors (1998a, 1998b); Gru È ner and Hefeker (1998); Cukierman and Lippi (1998,
1999); Soskice and Iversen (1999).
18. Arguing from a Barro±Gordon perspective, Calmfors (1998c) has conjectured that
incentives to reform inside the EMU are greater than outside, since countries with
control over monetary policy are likely to view labour market reform and monetary
policy as substitutes for reducing unemployment, while inside EMU the latter
vanishes. Reforming labour markets provides one means of ensuring against
idiosyncratic shocks. This effect will be strengthened by ®scal pressures stemming
from increasing unemployment, as well as the reorientation of national objective
functions when in¯ation can no longer be in¯uenced by national policies.
Similarly, Hefeker (1998) assumes unions which choose both the nominal wage
and the degree of ¯exibility.
19. For a discussion of these issues see Belke (1996).

20. See Bruno and Sachs (1985), Chapter 11, esp. pp. 232±40.
21. Willem Buiter (1995) has made this point, as have others. The results of Blanchard
and Katz (1992) imply adjustment to adverse shocks in the United States which are
long and drawn out over several years.
References
Agell, J. (1998) Òn the Bene®ts from Rigid Labour markets: Norms, Market Failures, and
Social Insurance', Uppsala Dept of Economics, Working Paper, 1998:17.
Akerlof, G. and J. L. Yellen (1985) À Near-rational Model of the Business Cycle, with
Wage and Price Inertia', Quarterly Journal of Economics, 100, 823±38.
Ball, L. and D. Romer (1990) `Real Rigidities and the non-Neutrality of Money', Review of
Economics and Statistics, 57, 183±203.
Barro, R. and D. Gordon (1983) À Positive Theory of Monetary Policy in a Natural Rate
Model', Journal of Political Economy, 91, 589±610.
Bayoumi, T. and B. Eichengreen (1993) `Shocking Aspects of European Monetary
Integration', in F. Torres F. and F. Giavazzi, Adjustment and Growth in the European
Monetary Union, Cambridge, Cambridge University Press.
ÐÐÐÐ (1996) Òperationalizing the Theory of Optimum Currency Areas', CEPR
Discussion Paper, 1484.
Bean, C. (1998) `The Interaction of Aggregate Demand Policies and Labor Market
Reform', Swedish Economic Policy Review 5, 353±82.
Belke, A. (1996) Ìmplikationen einer zentralisierten Geld- und WaÈhrungspolitik fuÈ r die
BeschaÈftigung in Europa', 8 UniversitaÈt Bochum, Diskussionspaper.
Berthold, N. and R. Fehn (1998) `Does EMU Promote Labor-Market Reforms?', Kyklos, 51,
509±36.
Bils, M. (1985) `Real Wages and the Business Cycle: Evidence from Panel Data', Journal
of Political Economy, 93, 666±89.
Blanchard, O. (1990) `Why does Money Affect Output: A Survey', Chapter 15 in
B. Friedman B. and F. Hahn (eds), Handbook of Monetary Economics, Amsterdam, Elsevier.
Blanchard, O. and L. Katz (1992) `Regional Evolutions', Brookings Papers on Economic
Activity, 1, 1±75.
Bo®nger, P. (1994) Ìs Europe an Optimal Currency Area?', CEPR Discussion Paper, 915.
Branson, W. and J. Rotemberg (1980) Ìnternational Adjustment with Wage Rigidity',
European Economic Review, 13 (3), 750±77.
Bruno, M. and J. Sachs (1985) Economics of Worldwide Stag¯ation, Cambridge, MA,
Harvard University Press.
Buiter, W. (1995) `Macroeconomic Policy during a Transition to Monetary Union', CEPR
Discussion Paper, 1222.
Burda, M. (1995) Ùnions and Wage Insurance', CEPR Discussion Paper, 1232.
Buti, M. and A. Sapir (1998) Economic Policy in the EMU, Oxford: Oxford University Press.
Calmfors, L. (1998a) `Macroeconomic Policy, Wage Setting and Employment ± What
Difference does EMU Make?', Oxford Review of Economic Policy, 14, 125±51.
ÐÐÐÐ (1998b) Ùnemployment, Labour-market Reform and Monetary Union',
Institute for International Economics Studies Seminar Paper, 639.
ÐÐÐÐ (1998c) `Monetary Union and Precautionary Labour-market Reform', Institute
for International Economics Studies Seminar Paper, 659.
Calmfors, L. and, Drif®ll (1988) `Bargaining Structure, Corporatism and Macroeconomic
Performance', Economic Policy, 6, 3±62.
Carleton, D. (1986) `The Rigidity of Prices', American Economic Review, 76, 637±58.
Cukierman, A. and F. Lippi (1998) `Central Bank Independence, Centralisation of Wage

Bargaining, In¯ation and Unemployment ± Theory and Evidence', CEPR Discussion
Paper, 1847.
Cukierman, Alex and Francesco Lippi (1999) `Labor Markets and Monetary Union: A
Preliminary Strategic Analysis', Tilburg University March mimeo.
Danthine, J.-P. and J. Hunt (1994) `Wage Bargaining Structure, Employment and
Economic Integration', Economic Journal, 104, 528±41.
DeÂcressin, J. and A. Fatas (1995) `Regional Labor Market Dynamics in Europe', European
DeGrauwe, P. (1991) Ìs Europe an Optimal Currency Area? Evidence from Regional
Data', CEPR, Discussion Paper DP, Number, 555.
Dohse, D. and C. Krieger-Boden (1998) WaÈhrungsunion und Arbeitsmarkt. Auftakt zu
unabdingbaren Reformen, Kieler Studien 290, Tu È bingen: Mohr.
Dornbusch, R. (1987) Èxchange Rates and Prices', American Economic Review, 77, 93±
106.
ÐÐÐÐ (1996) `The Effectiveness of Exchange Rate Changes', Oxford Review of Economic
Policy, 12, 26±38.
Dornbusch, R., C. Favero and F. Giavazzi (1998) Ìmmediate Challenges for the European
Central Bank', Economic Policy, 26, 15±64.
Duca, J. (1998) `How Increased Product Market Competition May Be Reshaping
America's Labor Markets', Federal Reserve Bank of Dallas Economic Review, Fourth
Quarter, 2±16.
Eichengreen, B. (1993) Èuropean Monetary Uni®cation', Journal of Economic Literature,
31, 1321±57.
ÐÐÐÐ (1998) Èuropean Monetary Union: A Tour d'Horizon', Oxford Review of
Economic Policy, 4, 24±40.
Fatas, A. (1997) ÈMU: Countries or Regions: Lessons from the EMS Experience', European
Feldstein, M. (1992) `The Case against EMU', The Economist, 13 June, 19±22.
Frankel, J. and A. Rose (1996) `The Endogeneity of the Optimum Currency Area Criteria',
CEPR Discussion Paper, 1473.
ÐÐÐÐ (1997) Ìs EMU More Justi®able ex post than ex ante?', European Economic Review,
41, 753±60.
Friedman, M. (1953) `The Case for Flexible Exchange Rates', in M. Fredman, Essays in
Positive Economics, Chicago, University of Chicago Press, 153±203.
Gros, D. and C. Hefeker (1998) `Factor Mobility, European Integration and Unemploy-
ment', Brussels, CEPS, mimeo.
GruÈ ner, H. and C. Hefeker (1999) `How will EMU Affect In¯ation and Unemployment in
Europe', Scandinavian Journal of Economics, 101, 33±47.
Hagen, J. von and M. J. M. Neumann (1994) `Real Exchange Rates Within and
Between Currency Areas: How Far Away is EMU?', Review of Economics and Statistics,
2, 236±44.
Hamermesh, D. (1993) Labour Demand, Princeton, Princeton University Press.
Hefeker, C. (1998) `Labour Market Rigidities, Unions and Monetary Union', University
of Bale, mimeo.
Jeanne, O. (1998) `Generating Real Persistent Effects of Monetary Shocks: How Much
Nominal Rigidity Do We Really Need?', European Economic Review, 42, 1009±32.
Kenan, P. (1969) `The Theory of Optimum Currency Areas: An Electic View', in
R. Mundell and A. Swoboda (eds), Problems of the International Economy, Chicago,
University of Chicago Press.
Knetter, M. (1989) `Price Discrimination by US and German Exporters', American

Kollmann, R. (1999) Èxplaining International Comovements of Output and Asset
Returns: The Role of Money and Nominal Rigidities', University Paris XII, June,
mimeo.
Mankiw, N. G. (1985) `Small Menu Cost and Large Business Cycles: A Macroeconomic
Model of Monopoly', Quarterly Journal of Economics, 100, 529±39.
McKinnon, R. (1963) Òptimal Currency Areas', American Economic Review, 53, 715±25.
Mundell, R. (1961) `The Theory of Optimal Currency Areas', American Economic Review,
51, 657±65.
Obstfeld, M. and Peri, G. (1998) `Regional Non-Adjusment and Fiscal Policy', Economic
Policy, 26: 205±59.
OECD (1994) The Job Study, Parts I and II, Paris, OECD.
Okun, A. (1981) Prices and Quantities A Macroeconomic Analysis, Washington:
Brookings Institution.
RoÈ ger, W. (1998) Òutput, Prices and Interest Rates Over the Business Cycle: How Well
Does a Keynesian±Neoclassical Synthesis Model Perform?', European Commission,
mimeo.
Romer, D. (1996) Advanced Macroeconomics, New York, McGraw Hill.
Sachs, J. (1979) `Wages, Pro®ts and Macroeconomic Adjustment: A Comparative Study',
Brookings Papers on Economic Activity, 2, 269±319.
ÐÐÐÐ (1983) `Real Wages and Unemployment in the OECD Countries', Brookings
Papers on Economic Activity, 1, 255±304.
Smith, J. (1999) `Nominal Wage Rigidity in the UK', Department of Economics,
University of Warwick, March, mimeo.
Soskice, D. and T. Iversen (1999) `Multiple Wage-Bargaining Systems in the single
European Currency Area', Oxford Review of Economic Policy 14, 110±123.
Streeck, W. (1998) `The Internationalisation of Industrial Relations in Europe: Prospects
and Problems', Working Paper Series in European Studies, 1.
Taylor, J. (1980) Àggregate Dynamics and Staggered Contracts', Journal of Political
Economy, 88, 1±24.
Wyplosz, C. (1997) ÈMU: Why and How it Might Happen', Journal of Economic
Perspectives, 4 (11), 3±21.
Discussion
Hendrik J. Brouwer
Almost everyone agrees that a certain degree of ¯exibilisation of Europe's labour

markets will be needed in order to help EMU absorb asymmetric shocks. We
don't know much about the probability that these shocks will occur in Europe.
What we do know is that the combination of EMU and the current wave of
product market liberalisation in Europe will make reallocations of production
factors increasingly likely. Production activities may become stronger concen-
trated along regional lines, as comparative advantages become more important
now that exchange rate uncertainty within Europe has been eliminated. We
also know that labour mobility is and will remain slower than in the United
States. It is therefore important that labour markets are able to accommodate
this process. Today's high unemployment in Europe points to the need of more
¯exible labour markets in order to reduce its structural component. The EU
Council has acknowledged this and it has put this issue high on the political
agenda. One could ask, however, whether it is also put high on the national
political agendas. It should be, since the ultimate responsibility rests with the
member states themselves. They are in the best position to determine which
speci®c measures should be taken for their own economies.
What types of labour market rigidities can we observe?
. A slow reaction of nominal wages to changes in real and/or nominal

conditions in the economy. This is the main rigidity addressed in Burda's
Chapter 7.
. But there is also evidence for lack of suf®cient wage differentiation ± for
instance, if wages are based on average productivity movements in the
economy. This causes a crowding-out of employment in low productivity
sectors, like parts of the service sector. In some countries too high a level of
legal or negotiated minimum wages reinforces this effect.
. Another phenomenon is the rather persistent high non-wage labour cost.
And, last but not least:
. Regulation and contracts limit a suf®cient degree of ¯exibility.
276
Discussion 277
Burda tells us that the call for political action to achieve greater ¯exibility does
not really matter much, since EMU itself will force labour markets in Europe to
become suf®ciently more competitive. His policy message is surprising and
indeed provocative. The question is whether that is the whole story. Let me
®rst note the issues on which I agree with Michael Burda:
. First, the chapter's central notion that EMU implies a major regime shift
which will not let the existing labour institutions in Europe remain intact.
. I also agree with the expectation that EMU may increase nominal price
rigidities. Not only the reasons mentioned in the chapter ± a reduced role
for exchange rate changes and for outside competition (pp. 258±61) ± could
play a role, but also the independent European Central Bank (ECB) with a
mandate of maintaining price stability will make a stable price climate in
Europe more likely. The general view that ®scal policy should contribute to
stability ± as expressed in the Stability and Growth Pact ± gives further
support.
. I share Burda's scepticism about the likelihood of centralised wage
negotiations in Europe (pp. 264±5). But there are some tendencies in that
direction. Therefore, I would like to go one step further: Is it really useful to
propagate and stimulate a European Social Dialogue? I doubt whether that
could enforce labour market ¯exibility. So far, the results of this dialogue
have produced the opposite.
As I said, however, I am not convinced about all elements of Burda's chapter.

Let me mention where I have some dif®culties:
. The chapter states that real rigidities will diminish through more
competition in the labour market which will lead to weakening of union
power in wage determination (p. 264). This contrasts with the recent high
wage demands in Germany, which do not give the impression of
moderation on the side of the unions brought forth by EMU.
. Of course, one could argue that such high wage demands are transitory but
that still leaves the question of what time frame Burda has in mind with
regard to this change in rigidities. In the meantime, when moving from one
state of equilibrium to another, things could still go wrong. Hysteresis
effects explain why Europe has such a high level of structural
unemployment.
. What could go wrong is not only that wage demands are excessive, but also
that governments resist the transfer of a part of their coordination role to
the market. On the contrary, painful adjustment in the labour market may
well lead to a call for increased coordination by governments in the
direction of short-term `quick ®xes', like the 35-hour working week, while
neglecting the need for sometimes unpopular structural adjustment.
278 Hendrik J. Brouwer
. I think the chapter somewhat underestimates the role governments have to

play, as this element is hardly discussed at all. It is well documented that
current rigidities in the European labour markets relate not just to wage
determination but also to laws and regulations, such as rules on working
hours and on hiring and ®ring. There is much room for improvement and
thus for action by politics in these areas.
. I think we cannot exclude that a quite different and less rosy picture than
the one sketched by Burda would result if nominal wages tended to exhibit
the same degree of stickiness as prices. I see some reasons for such a
development, such as (1) a possible lengthening of the duration of wage
agreements and (2) a ¯oor in nominal wage contracts in a situation of low
in¯ation, which can already be observed in some countries.
. Finally, the empirical outcomes which the chapter presents are somewhat
puzzling. It seems that some trends of the last decades already display the
direction to which Burda's hypotheses point. For instance, it is noted
(p. 267) that Tables 7.7 and 7.8 show an increasingly uncorrelated real wage
behaviour in the EU over time. This prompts me to ask why these trends
have already been operating for so long. Should this not be taken as
indication that other variables than the coming of EMU have played a role
in this respect? Also, the robustness of some of the outcomes is
questionable. For instance, the correlations presented in Tables 7.2 and
7.3 (pp. 260±1), which measure the synchronisation of in¯ation, seem to
contradict each other. According to Table 7.2 the degree of synchronisation
of in¯ation has remained high, especially in the EU core group, whereas
Table 7.3 suggests that this synchronisation has fallen all over the EU.
The conclusion of my comments is that, although enticing, it is better not to

believe too much that the ìnvisible hand' of monetary integration will clear
labour markets, but rather focus on how policy makers can help reduce
existing rigidities. So I would like to change the sentence on p. 270: `Calls for
additional ¯exibility may be the economic equivalent of whipping an
apparently dead horse', and that could be a very useful thing to do!
Discussion
LuõÂs Campos e Cunha
1 Introduction
Burda's chapter 4 is not an easy chapter to comment on. Its aim is `to entertain
new-Keynesian arguments' without a well speci®ed model and, therefore, no
clear policy recommendation can be derived from it, despite being related to
policy issues. Furthermore, it is also not an empirical paper, because `the
empirical evidence is rather modest' and it is questionable that such evidence
would be of any relevance owing to the change of regime, as correctly
emphasised by Burda. The chapter deals with institutional and sociological
matters for discussing the impact of the change of economic regime due to
monetary union, with particular emphasis on labour markets. As Burda says,
ìs highly speculative, but meant to be so'. I would say it is both speculative and
provocative, and this is an excellent idea.
Given these dif®culties, let me summarise my understanding of the main
issues raised by the chapter and make some comments as I go along.
2 EMU and OCA: Mundellian problems
Some of the issues the chapter promises to address are those found in the
literature on optimal currency areas (OCA) following Mundell (1961). What is
the optimal response of monetary policy to idiosyncratic shocks in an
economy? Should there be one or multiple currencies, in order to respond
optimally to these shocks? As the literature stands now, we are not in a
condition to answer all these questions satisfactorily. However, one thing is
certain: the Mundellian nightmare regarding OCA in the context of EMU has
been oversold. The understanding of exchange rate policy is very different
today than it was in 1961. More speci®cally, it is today widely accepted that no
exchange rate policy can insulate an economy from real shocks. To assert that
exchange rates can do more than insulate a country from nominal shocks is an
incorrect reading of Mundell (1961). This clari®cation has been recently made
in Mundell (1997, 1998).
279
280 LuõÂs Campos e Cunha
The lack of factor mobility ± namely, labour ± is another related problem

raised in Burda's chapter. I do not think that the lack of actual signi®cant
labour ¯ows within the EU-11 is a sound proof of rigidities in this area, as
claimed by Burda. First, as already noted, due to both factor price equalisation
(FPE) and to capital ¯ows. These have been large and increasing during the last
decade and, in a certain class of models, they are perfect substitutes for labour
mobility. Secondly, the ùnwillingness' of people to move, explained by older
demographics as referred to by Burda, is not a distortion. If people do not want
to move, why should they? Unwillingness to migrate across borders is not a
component of rigidity or a distortion of any sort.
As a ®nal note, in the 1960s, there was large labour migration within Europe
despite its being much more dif®cult than today from any point of view.
When people were willing to move, they did.
3 EMU: a change of regime
A second cornerstone of Burda's chapter is the change of regime, in the sense

used in Lucas (1976). This is a very important and relevant issue and its impact
both in the short run and long run is dif®cult to assess. One should note that
an argument of change of regime is an elegant one, but it remains to be seen
whether it is relevant as concerns factor market rigidities in Èuroland'. In any
case, in the chapter it is argued that nominal rigidities will increase but real
wage ¯exibility will prevail.
Real wage ¯exibility

As regards wages, it is claimed that EMU will lead to a change in institutions of
wage determination that will make the EU look more like the United States
today, in the sense that the change of regime will imply an increase of nominal
rigidities and a reduction of real rigidities.
The argument is based on the idea that trade unions will lose much of their
bargaining power and, therefore, nominal and real wage will become more
¯exible. On the one hand, it is not clear why national trade unions should be
weakened. Alternatively, one could argue that competition among them will
increase, leading to a reversed conclusion. On the other hand, most of the
rigidities of the labour market mentioned are due to regulation: I have in mind
minimum wage laws, unemployment compensation schemes, restrictions to
®rms to adjust their labour force, and so on. These will not vanish either because
of monetary union or the lack of international coordination of national labour
unions, for the same reason as, in the past, those rigidities were not a consequence
of the existence of several currencies or of labour union coordination.
Wage bargaining and the role of labour unions under the new regime could
be better discussed in a model of insiders/outsiders. We should then look for
changes that could induce an increase or decrease in the reaction of wages to
local employment conditions and to unemployment persistence.
Discussion 281
Price rigidities
The chapter also argues that as the regime changes, price stickiness in the
goods market will increase. The author claims that EMU will lead to an
increase in nominal rigidities because the share of `home goods', goods traded
inside Europe, is larger than the total of home goods across all countries in the
Union. The reasoning is that these goods will not be exposed to ¯uctuations
and this will favour nominal rigidities. Both parts of the argument need
clari®cation: is Burda arguing that shocks to goods traded inside the EU were
mainly driven by national monetary policies? Otherwise, ¯uctuations in these
sectors will certainly persist. Furthermore, why should a more stable
environment lead to higher nominal rigidities?
This ìnwardisation' effect owing to the single currency is probably too
much emphasised, since most of the effects were anticipated at least for the so-
called `core' countries. Furthermore, the `closing' of the Euroland economies
does not have the same nature of a country imposing tariff barriers and capital
controls. With the single currency the domestic economies become bigger,
and there is an increase in competition owing to reduced transaction costs and
increased price transparency. Market segmentation is less likely to take place
and, therefore, ®rms are more ± not less ± exposed to competition as economic
frontiers are pushed further away, so it is not obvious that monopolistic power
will necessarily increase.
4 Concluding remarks on policy
I would like to believe, using the wording of the chapter, that, ÈMU is the
Trojan horse of decentralisation' and that the Èuro is the Trojan horse which
will liberalise labour markets', and therefore, `calls for additional ¯exibility
may be the economic equivalent of whipping a dead horse'. Unfortunately,
most of us feel that we have to continue to call for some structural reforms
leading to institutional and legal changes so that structural unemployment
will fall. The Euro will not be the solution for high unemployment rates.
Unlike Burda, I am more concerned with the symmetric consequences in the
new regime. We could imagine consequences concerning labour markets
owing to compulsory centralised harmonisation. In fact, regions of Euroland
with lower labour productivity should be vigilant and oppose attempts to
harmonise labour market regulation throughout the Union. Just imagine the
impact of a forceful and a speedy harmonisation of minimum wage towards
the core country levels. It would be the equivalent of Brussels offering a rope to
thousands of workers to hang themselves and it would be very dif®cult for
them to resist. This would imply a jump in unemployment rate for lower-
productivity regions as real rigidities increased. This is foreseeable and it would
be a big step in the wrong direction.
As regards monetary policy, Burda makes a point that stronger price
persistence in Euroland will facilitate a more powerful monetary policy, which
282 LuõÂs Campos e Cunha
is almost an invitation to a more active monetary policy. Without a more

detailed mechanism of how persistence comes about, this is a claim dif®cult to
make and at the same time to dispute. For instance, we cannot talk about
`monetary policy effectiveness' without saying how price expectations are
formed, which is not explained in the chapter.
On the other hand, if one believes ± as I do ± that money is basically neutral
in the long run, an increase in price persistency, if true, would call for the
European Central Bank (ECB) to have an even longer-run orientation while
conducting its monetary policy. A mistake made today will be paid further
away in time owing to the increased persistence of prices in Euroland: the long
lags of monetary policy will become even longer. To sum up, persistency does
not necessarily facilitate the job of the ECB.
I would like to ®nish by stressing that the chapter makes a valuable
contribution to the debate and it was a pleasure to participate in it.
References
Lucas, R. (1976), Èconometric Policy Evaluation: A critique', in K. Brunner and
A. H. Meltzer (eds), The Phillips Curve and Labor Markets, Amsterdam, North-Holland,
19±46.
Mundell, R. (1961), `The Theory of Optimal Currency Areas', American Economic Review,
51, 657±65.
ÐÐÐÐ (1997) `Currency Areas, Common Currencies and EMU', American Economic
Review, (2) 214±16.
ÐÐÐÐ (1998) `Great Expectation for the Euro', The Wall Street Journal, March, 24±5.
8
The Monetary Transmission Process:

Concluding Remarks1
Otmar Issing
1 Introduction
In order to be successful in conducting monetary policy, central banks need to

have a good understanding of the working of the economy, including an
accurate assessment of the timing and the effects of changes in the policy
instrument on in¯ation and economic activity ± that is, the monetary
transmission process (MTP). Such an assessment is necessary in order to tailor
the policy response to unexpected developments in the economy and
successfully maintain price stability.
The introduction of the Euro constitutes a unique experience in monetary
economics and central banking. This historical regime change is likely to affect
the size and the timing of the various transmission channels. Analysing the
potential changes in the structure of the economy caused by the start of Stage
III of EMU is therefore of great importance and will surely be an object of
research for decades to come. Economists are indeed likely to regard the
introduction of the Euro as a `laboratory experiment' of the Lucas Critique.
However, the Eurosystem does not have the luxury of waiting for the results
of this research. The Governing Council started conducting an independent
monetary policy on 4 January 1999. Since then it has met twice a month to
assess the outlook for in¯ation and decide on the appropriate policy actions.
Central bankers are used to making decisions on the basis of incomplete,
evolving information. As you all know, it was this lack of precise knowledge
about the variable and long transmission lags of monetary policy that led
Milton Friedman to warn against an activist monetary policy.
The stability-oriented strategy announced by the Council in October 1998 is
designed to guide monetary policy through the uncertainties that must
inevitably accompany the transition to the single currency. Of course, it is in
no way a substitute for a thorough analysis of the state of the Euro Area
economy, its implications for the in¯ation outlook and the policy actions
necessary to achieve price stability. On the contrary, the implementation of
the strategy presupposes a continuing investigation of the working of the Euro
283
284 Otmar Issing
Area economy, including the transmission process. In this respect, confer-

ences like this one are a very important source of information, particularly if
they bring together such a distinguished group of outstanding researchers on
monetary economics. Let me then use this opportunity to brie¯y give you my
own reading of the key messages of the discussion.
I would like to review two key aspects. First, what do we know about the
MTP in the Euro Area and, second, how it is likely to change?
2 What do we know about the MTP in the Euro Area?
There is by now quite a large literature on the MTP in the Euro Area countries.
While the focus of this literature has often been on the cross-country
differences and the implications for the common monetary policy, a lot can
also be learned from this literature concerning Euro Area-wide transmission.
Two approaches have been used which I will discuss in turn. One approach is
to describe and analyse the institutional features of the economies which have
a bearing on the MTP. Another is to use econometric models to estimate the
effects of a change in policy-controlled interest rates on economic activity and
in¯ation. Let me brie¯y review the results of both approaches.
Institutional comparison
The way in which changes in monetary policy feed through the various sectors
of the economy depends to a large extent on the structural and institutional
features of its ®nancial, labour and goods markets. These institutions are
rooted in the history of each country and often resulted from the interplay
between particular macroeconomic shocks, the preferences and bargaining
powers of various agents in the economy and more or less market-friendly
regulatory and government forces.
Researchers have devoted a lot of energy to investigating whether such
institutional characteristics, which from a theoretical perspective may affect
the impact of monetary policy, differ across countries or regions. Often the
focus on one particular aspect of the MTP leads to the conclusion that the
transmission of monetary policy impulses will be very heterogeneous across
Euro Area countries.
However, if one brings several of these features together (as is done in
Table C1.1), then it is much less obvious that the effects of policy will be very
different. Generally, countries that are particularly sensitive to policy changes
because of one criterion, will be less sensitive on the basis of other
characteristics. Let me brie¯y go through some examples.
Goods markets
One important characteristic of the goods markets which affects the transmis-
sion channels of monetary policy is the degree of openness. While the Euro
Area as a whole will be a relatively closed economy ± with a ratio of exports to
Concluding Remarks 285
GDP of about 15 per cent, it is only marginally more open than the United
States or Japan ± there is a fair degree of dispersion within the Euro Area (see
Table C1.1). For example, Belgium is about three times as open to the non-
Euro Area as Spain. As a consequence, while the exchange rate channel may be
relatively less important for the Euro Area as a whole, the strength of its impact
on output and in¯ation may be quite different across countries.
Similarly, the overall interest elasticity of aggregate demand will depend on
the importance of interest rate sensitive sectors in the economy. For example,
Carlino and DeFina (1998) ®nd that, within the United States, regions with a
higher weight of construction and manufacturing in the economy are more
strongly affected by the Federal Reserve System's monetary policy. If similar
considerations are true within the Euro Area, then one could expect that, for
example, Germany and Spain should experience a stronger interest rate
channel than the Netherlands or Belgium.
Another sector which is highly responsive to the level of the interest rate
is the housing sector. Maclennan, Muellbauer and Stephens (1998) have
expressed concern about the implications of cross-country differences in the
tax and legal framework of housing markets for the transmission mechanism.
In particular, transaction costs owing to taxes on new houses and stamp duties
are quite heterogeneous across countries. A particularly low tax rate in the
United Kingdom is consistent with the high amplitude of its housing market
cycle. Within the Euro Area, Germany, Italy and Spain have much smaller
transaction costs than the Netherlands, France and Belgium.
Financial structure
Table C1.1 also summarises a number of features of ®nancial structure across
countries. From a comparison of these features across countries, a number of
tentative conclusions can be drawn. Countries in the Euro Area appear to have
several features in common. Stock markets and corporate debt securities
markets are less well developed than in the United States, the United Kingdom
or even in Japan. Of the seven largest Euro countries, the Netherlands is the
only one where the number of publicly traded ®rms and the stock market
capitalisation show levels that are comparable to those of the United States or
the United Kingdom. Obviously, this implies that equity prices in the Euro
Area should play a less important role in the MTP through both the effect on
the cost of capital and wealth effects on consumption.
Similarly, France is the only country where the corporate sector issues a
signi®cant share of its debt directly on the market. In 1997, up to one-third of
the debt of French ®rms consisted of securities, while it was below 10 per cent
in the other countries. To the extent that bank lending rates are more sticky
than market rates in response to a change in policy-controlled rates, the
interest rate channel should be weaker in the Euro Area.2
However, there remain some important differences in ®nancial structure
within the Euro Area. For example, the degree of ®nancial intermediation is
Table C1.1 Institutional features with a bearing on the transmission mechanisma
Goods markets Financial markets Labour markets
Non-Euro Manufactur. Stamp duty Number of Stock Securities Credit at Union Employment Unemploy.
exports and VAT publicly market / Total ®rm adjustable coverage protection bene®t
/ GDP construction on new traded ®rms capitalisation debt rate index index index
(per cent) / GDP homes / per capita / GDP (per cent) / GDP
(per cent) (per cent) (per cent) (per cent)
Belgium 26.8 27 33 13.7 39 7

38 3 17 240
Germany 14.0 37 2 8.3 24 6
49 3 15 252
Spain 8.1 32 9 9.1 35 9
34 / 50 3 19 245
France 10.5 27 25.6 11.7 34 15
40 3 14 171
Italy 11.7 30 8.2 3.8 19 5
47 3 20 10
Netherlands 18.5 24 23.5 14.0 90 3
29 3 9 140
Austria 12.9 29 13.1 14 2
65 3 16 100
Average*b 14.6 29 16.9 11.3 36 6.7 44.7 3 15.7 165.4
USA 8.5 31.9 95 20 39 1 1 25

Japan 10 18.6 72 10 71 2 8 30
UK 21.4 1 41.4 127 19 85 2 7 152
Year 1997 1980±90 1997 1996 1995 1993 1 993 1989±94 1989±94 1989±94
Source IMF, Carlino and Maclennan, Cecchetti Prati and Borio Borio Nickell Nickell Nickell
Direction of DeFina Muellbauer and (1999) Shinasi (1995) (1995) (1997) (1997) (1997)
Trade (1998) Stephens (1997)
Statistics (1998)
Notes: a These indicators should be taken as suggestive.

b
Simple average of the seven countries above.

Goods markets indicators: Share (in GDP) of exports vis-a-vis countries that are not in the Euro Area. For the United States, Japan and the United Kingdom: total exports; Share of
manufacturing and construction sectors in GDP; Sum of stamp duty on housing transactions and VAT rate on new houses.
Financial markets indicators: Number of publicly traded ®rms per million of population; Stock market capitalisation scaled by GDP; Share of securities debt in non-®nancial corporate
sector total debt; Credit at adjustable rate scaled by GDP ± i.e. total credit over GDP times the proportion of total credit at adjustable rate as in Borio (1995).
286
quite different among the large Euro Area economies with loans of credit
institutions to non-banks ranging from as low as 67 per cent of GDP in Italy, to
as high as about 150 per cent of GDP in Germany and in the Netherlands.
Similarly, the maturity structure of the debt of the private sector differs
considerably across countries (Borio, 1995).
Labour market structure

The wage-setting process is a key element in the transmission from monetary
policy impulses to prices. Continental European countries are generally
characterised as having lower nominal wage rigidity but higher real wage
rigidity than the US.3 The implications for the transmission process are not
unambiguous. Higher real wage rigidity in the Euro Area implies that the
output cost of achieving a similar effect on in¯ation may be greater. This
would tend to increase the sacri®ce ratio. On the other hand, a high response
of nominal wages to expected in¯ation has the potential of making the direct
transmission process of monetary policy on wages and prices much more
effective to the extent that the central bank's in¯ation objectives are highly
credible.
A comparison of labour market structures across countries has usually put
more emphasis on understanding the level of European structural unemploy-
ment than the transmission process.4 Except for the United Kingdom, where
the labour market is generally much less regulated than on the continent, it is
dif®cult to rank countries because there are many dimensions of labour market
rigidities.5
Euro Area labour market institutions are generally characterised by
unemployment bene®ts, worker protection and union participation that are
much larger than in the United States. Yet, the hierarchy among Euro
countries in terms of institutional factors of rigidity is far from obvious.
Although most Euro countries have a high degree of centralisation in wage
negotiations, there are marked differences along some other dimensions. For
example, when it comes to employment protection, Italy and Spain appear to
be the most rigid, while the Netherlands is much less so. In contrast, in terms
of the generosity and duration of unemployment bene®ts, Italy is far below
other Euro countries, while Germany, Spain and Belgium stand at the top.
In sum, this brief overview of institutional features shows that in many
respects the MTP in Euro Area countries share several common features.
Moreover, although important institutional differences remain, the ranking of
countries in terms of their implications for the strength of the MTP is not clear.
Countries that score low on one criterion may score high on another. Indeed,
the average ranking of the seven biggest countries in the Euro Area according to
the qualitative impact associated to each of the ten indicators of Table C1.1
suggests there are no clear differences in the strength of the MTP among these
countries.
288 Otmar Issing
Macroeconometric analyses of the MTP

This general picture is con®rmed by the many studies that have examined the
transmission process in Euro Area countries using a variety of macroecono-
metric approaches (vector autoregressions (VARs), large-scale macroecono-
metric models, smaller-scale calibrated models, etc). As emphasised by Kieler
and Saarenheimo (1998), there are no cross-country differences in the strength
and timing of the MTP that are robust across these studies. While this result
may in part be due to differences in methodologies and the statistical
imprecision of some of the estimates, it suggests that it is not straightforward
to ®nd clear cross-country differences. In light of our ®ndings concerning the
institutional differences across countries, this is not too surprising.
Overall, the response of real GDP and in¯ation to a monetary policy
tightening follows the pattern predicted by textbook theory. An increase in the
policy-controlled interest rate leads to an immediate decline in GDP, which
reaches its trough after ®ve±eight quarters before it comes back to the base-line.
The combined impact of a decrease in import prices following the appreciation
of the exchange rate and downward pressure on wages coming from rising
unemployment leads to a fall in prices. Typically, however, the fall in in¯ation
takes somewhat longer to materialise than that of output. This is also the
pattern of responses that can be obtained from simulating the effects of a
monetary policy tightening in various models estimated on aggregate Euro Area
data. Table C1.2 shows the effects of a policy tightening in the Euro Area using
three different macroeconometric approaches: a VAR model by Monticelli and
Tristani (1998), a two-equation model estimated by Peersman and Smets (1999)
and the ECB's area-wide model. There are many reasons why models estimated
on synthetic historical data may be biased and give the wrong answers.
Nevertheless, on the basis of existing evidence two conclusions can be drawn.
First, the qualitative effects are very similar to what I have previously described
and what has been found for the United States. Second, there is considerable
uncertainty about the precise timing and size of the effects on economic activity
and in¯ation. Let me now turn to the question of how the MTP and differences
across countries may be affected by the creation of the single currency.
3 The monetary transmission process in the Euro Area: is it likely

to change?
For the Eurosystem it will be even more important than for most other central
banks to continuously monitor any developments that may change the
transmission process of monetary policy. I would like to focus on the likely
developments in the ®nancial sector and its impact on the MTP.
I will not elaborate on the potential effects of the single currency on goods
and labour markets. Let me just underline that, particularly in the case of
labour markets, further structural reform is needed not only to reduce the
Table C1.2 Simulation of a monetary policy tightening
Monticelli and Tristani (1998) Peersman and Smets (1999) ECB area-wide model
Horizon Year 1 Year 2 Year 3 Year 1 Year 2 Year 3 Year 1 Year 2 Year 3
Interest rate 0.102 0.044 ±0.025 0.5 0.5 0 0.5 0.5 0.5
GDP ±0.18 ±0.4 ±0.38 ±0.18 ±0.32 ±0.27 ±0.07 ±0.16 ±0.23
Consumer prices 0.07 ±0.16 ±0.45 ±0.11 ±0.7 ±1.8 ±0.11 ±0.22 ±0.25
289
290 Otmar Issing
equilibrium unemployment rate but also to increase labour market ¯exibility

and thereby improve the ability of the economy to cope with macroeconomic
shocks. Our hope is that the establishment of a stability-oriented single
monetary policy in the Euro Area will provide incentives to continue on the
path of reform.
At this point, I would also like to stress that, given the historically
unprecedented nature of EMU, it is natural to emphasise the potential
importance of structural breaks owing to the start of Stage III. One should,
however, realise that, ®rst, a long and gradual process of monetary
convergence has preceded the introduction of the new currency, so that
many of the adjustments may already have taken place. Second, the effect of
the single currency on competition in goods, labour and ®nancial markets
follows previous structural changes such as the Single Market initiative, so that
whatever structural change will take place is part of an on-going process.
The single monetary policy

As of 4 January 1999 the Eurosystem started to implement a single monetary
policy for the Euro Area. This has by itself potentially large effects on the way
changes in policy instruments affect the economy. In fact, it is generally
meaningless to discuss the MTP without a discussion of the monetary policy
regime. The monetary policy strategy and its credibility will affect the way
changes in policy instruments are interpreted by the private sector and feed
through into their expectations and actions. Let me brie¯y give two examples.
The effects of the single monetary policy on the MTP will be most visible in
the way asset prices respond to policy. For example, while in the past
responses of long-term interest rates to policy innovations were quite diverse
across Euro Area countries, more recently there have been abundant signs
of the development of a single private sector yield curve, which responds
homogeneously irrespective of the nationality of the borrower.
Second, the credibility of the Eurosystem's price stability objective is likely
to promote longer-term nominal contracting in both labour and ®nancial
markets. Such developments are again likely to homogenise the transmission
process. For example, evidence suggests that in Italy, where historically the
average maturity of debt contracts was relatively short owing to high and
variable in¯ation, an increasing proportion of the new housing debt is issued
at a ®xed interest rate.
The emergence of an integrated ®nancial market in the Euro Area

Monetary integration is a major step in the completion of the integration of
European ®nancial markets. Such an integrated ®nancial market denominated
in Euros may provide investors and issuers with levels of deepness and
liquidity which they could ®nd before only in the United States.
These developments will be accompanied by the spreading of ®nancial
market innovation across Europe. For example, direct ®nance through
commercial paper and corporate bonds for non-®nancial ®rms may increase in
countries where it has not yet developed and reach levels that are currently
observed only in France (a third of the corporate sector debt). The potential
consequences of the development of direct ®nance and a changing role of the
banking sector in Europe for the transmission may be signi®cant. Competition
between direct ®nance and bank ®nance will put pressure on banks and may
increase the speed of transmission from the money market rate to the rate at
which ®rms borrow. For example, in France, evidence shows that the money
market rate has taken over the prime rate as the reference to compute the bank
credit rate to big ®rms as well as to smaller ones. In the short run, it may,
however, also increase the fragility of the banking sector.
On the other hand, cultural preferences for ®xed rate contracts and against
indebtedness, national consumer protection laws and inertia in consumer
habits will tend to slow down cross-country competition in retail banking.
Retail banking involves long-term investment in brand names, legal expertise
in speci®c national laws and monitoring of borrowers' solvency. These provide
incumbent banks with substantial advantages over competitors, especially
foreign ones, so that initially the impact of EMU on retail banking could be
limited to the harmonisation of banks' funding costs.
The development of the Euro as an international currency

The development of the Euro as a strong and credible currency traded in deep
and liquid ®nancial markets may also affect the importance of the exchange
rate channel to the extent that it enhances the role of the Euro as an invoice
and anchor currency. While currently the exchange rate channel may be
somewhat more important than in the United States, its importance may be
further reduced if several trade partners ± for example, in Eastern Europe ± peg
their currency to the Euro and to the extent that the price of some
commodities could be traded in Euros.
4 Conclusion
This review of the existing literature on the MTP in the Euro Area con®rms the
validity of Friedman's famous dictum that monetary policy lags are long and
variable ± and, I would add, very uncertain. In the case of the Euro Area,
uncertainties are magni®ed by the scarcity of reliable aggregate data and the
potential effects of a regime change. The monetary policy strategy of the
Eurosystem was designed to face, in a pragmatic, yet conceptually sound way,
the many uncertainties inherent in this regime change.
The strategy is at the same time transparent and ¯exible. It clearly
communicates the commitment to price stability by de®ning price stability
as an increase of the area-wide harmonised index of consumer prices of below
2 per cent. This helps to anchor in¯ation expectations and preserve the anti-
in¯ationary reputation inherited from its precursors, the national central
292 Otmar Issing
banks (NCBs). In addition, two pillars ± a reference value for the growth of a
broad money aggregate and a broad-based assessment of the outlook for
in¯ation ± are used to explain monetary policy decisions necessary to achieve
the price stability objective. The prominent role of money ± which we consider
as a reference value, not a target in the traditional sense intended in the
monetary literature ± underlines the principle of continuity and is rooted in
robust theoretical and empirical arguments accumulated over many decades
of research. Indeed, probably the only statement with which everybody would
agree is that in¯ation is, in the long run, a monetary phenomenon. However,
uncertainties in the stability of money demand would make it inappropriate to
automatically respond to deviations of the growth in money aggregates from
target. This is why the Council decided not to adopt a strict monetary targeting
framework. Instead, the prominent role of money as a reference value
is complemented by a broad-based assessment of the in¯ation outlook,
including both model-based forecasts and a variety of other relevant
indicators. Together these two pillars allow for a timely and ¯exible response
to a changing environment, while keeping the objective of price stability in
clear focus.
The strategy will help communicating with the public and organising policy
action in a coherent way. We do not see it as an invariable dogma. The ECB
closely follows the developments in economic research: this is a key factor in
our continuous assessment of the strategy itself. We look forward to a
continuous, productive exchange between the ECB and the academic world
on these issues.
Notes
1. I thank Frank Smets and BenoõÃt Mojon for their valuable assistance.
2. There remains uncertainty about the effect of bank dependence on the MTP. For
example Cecchetti (Chapter 5 in this volume) has emphasised that countries in
which ®rms are more bank-dependent will be more sensitive to the Eurosystem's
decisions to change interest rates through a kind of credit channel. However,
Schmidt has argued in chapter 6 in this volume that long-term relationship bank
lending, which occurs more naturally in a bank-dominated ®nancial system, has led
banks to insure customers against interest rate shocks.
3. See, for example, Calmfors (1998).
4. See, for example, Blanchard (1998) or Bean (1994).
5. In addition, Burda's Chapter 7 in this volume shows that there is not yet a consensus
as to how the institutional features of continental European labour markets
in¯uence macroeconomic outcomes.
References
Bean, C. (1994) Èuropean Unemployment: A Survey', Journal of Economic Literature,
32(2), 573±619.
Blanchard, O. (1998) Èuropean Unemployment. Shocks and Institutions', Baf® Lecture,
mimeo.
Borio, C. E. V. (1995) `The Structure of Credit to the Non-government Sector and the
Transmission Mechanism of Monetary Policy: A Cross-country Comparison', in Bank
for International Settlements, Financial Structure and the Monetary Policy Transmission
Mechanism, CB 394, Basle.
Burda, M. (1999) Èuropean Labour Markets and the Euro: How Much Flexibility Do We
Really Need?', Chapter 7 in this volume.
Calmfors, L. (1998) `Macroeconomic Policy, Wage Setting and Employment ± What
Difference Does EMU Make?', Oxford Review of Economic Policy, 14, 125±51.
Carlino G. and R. DeFina (1998) `Monetary Policy and the U.S. States and Regions: Some
Implications for European Monetary Union', Federal Reserve Bank of Philadelphia,
Working Paper, 98±17.
Cecchetti, S. (1999) `Legal Structure, Financial Structure and the Monetary Policy
Transmission Mechanism', Chapter 5 in this volume.
Kieler M. and T. Saarenheimo (1998) `Differences in Monetary Policy Transmission?: A
Case not Closed', Economic Papers, 132, European Commission, Directorate-General
for Economic and Financial Affairs.
Maclennan, D., J. Muellbauer and M. Stephens, (1998) Àsymmetries in Housing and
Financial Markets Institutions and EMU', Oxford Review of Economic Policy, 14, 54±81.
Monticelli, C. and O. Tristani (1998) `What Does the Single Monetary Policy Do? A SVAR
Benchmark for the European Central Bank', DG Research of the European Central
Bank, mimeo.
Nickell, S. (1997) Ùnemployment and Labour Market Rigidities: Europe versus North
America', Journal of Economic Perspectives, 11(3), 55±74.
Peersman G. and F. Smets (1999) `The Taylor Rule: A Useful Monetary Policy Benchmark
for the Euro Area?', International Finance, 2(1), 85±116.
Prati, A. and G. Shinasi (1997) ÈMU and International Capital Markets', in P. Masson,
T. Krueger and B. Turtelboom (eds), EMU and the International Monetary System,
Washington, DC, International Monetary Fund.
Schmidt, R. `Differences between Financial Systems in European Countries: Conse-
quences for EMU', Chapter 6 in this volume.
Francesco Giavazzi
The wave of mergers and acquisitions that is sweeping the European banking
industry should not be a matter of indifference for those concerned with the
possibility that the monetary transmission mechanism may work asymme-
trically across EMU: for two reasons. First, banks, as is well known, are at the
centre of ®nancial markets in continental Europe, and thus of the transmis-
sion mechanism. Second, the consolidation of the European banking
industry, which is currently happening, may sharpen, rather than reduce,
the existing sources of crosscountry asymmetries.
Evidence of the importance of banks in the transmission mechanism inside
EMU is related to the extent to which monetary policy operates through a
`credit' channel, rather than simply through a `money' channel. As is well
known, when loans and bonds are imperfect substitutes in the balance sheets
of banks, following a squeeze in liquidity, banks reduce the amount of loans
they supply. Firms could turn to the bond market, but if bonds and loans are
imperfect substitutes the external ®nance premium will go up. This effect
[known as the `credit' channel (see, for example, Bernanke and Gertler, 1995)]
works on the supply side and ampli®es the more traditional demand effect of a
monetary tightening ± the change in interest rates which affects new marginal
spending by modifying borrowing conditions and by affecting asset prices,
and thus the market value of wealth.
Because of the large share of bank loans in the total debt liabilities of
European non-®nancial ®rms (85 per cent in Germany, 80 per cent in France
and Spain, 95 per cent in Italy, as opposed to 32 per cent in the United States)
the behaviour of banks is thus central to understanding the transmission
mechanism in Europe.
Reliable identi®cation of a `credit' channel requires the use of micro data:
macroeconomic time series are ill-suited to identify a `credit' channel from a
`money' channel in the transmission of monetary policy from the central bank
to banks. This is because the money channel works through banks' liabilities,
while the credit channel works through their assets, but assets and liabilities
are tightly related by accounting identities. For this reason, the evidence
proposed by macroeconomic studies which look at output and price
¯uctuations in response to shifts in the quantities of loans and deposits is
rarely decisive. On the contrary, microeconomic data allow one to ask whether
the responses of banks and ®rms to a shift in monetary policy differ depending
on their characteristics ± their size, for instance: small ®rms are more likely to
be liquidity-constrained and to depend on banks for ®nancing; similarly, small
banks ®nd it more dif®cult to insulate their loans' portfolio from a squeeze in
central bank liquidity, because a small bank typically cannot use bond
294
holdings as a buffer (Kashyap and Stein, 1997; Kashyap, Stein and Wilcox,
1993). So far, the empirical evidence on the importance of bank characteristics
in determining the response of loans to a shift in monetary policy has been
limited to US data. Favero, Flabbi and Giavazzi (1999) have extended this
evidence to Europe.
Their ®ndings, albeit limited to a speci®c episode ± the EMU-wide monetary
contraction which occurred during 1992 ± point in two main directions. In
Germany there is clear evidence of a lending channel for all banks, except the
smallest ones ± that is, for over 80 per cent of the German banking sector. In
France, on the contrary, there is no evidence of a lending channel,
independently of the size of the banks considered. The large Italian banks
are similar to the corresponding German banks, in the sense that there is
evidence of a lending channel: the impact on the supply of loans of a change
in reserves is however twice as large as for the corresponding German banks.
The difference between Germany and Italy lies in medium-sized banks (whose
share of the domestic banking industry is larger in Italy than it is in Germany):
while there is evidence of a lending channel for medium-sized German banks,
this is not the case for the corresponding Italian banks. In Spain, there is
evidence of a lending channel for medium-sized banks: the impact on lending
of a cut in reserves is similar to that found in Germany.
The empirical evidence thus points to signi®cant cross-sectional and cross-
country differences in the response of individual banks to monetary policy ± at
least in the speci®c episode studied in Favero, Flabbi and Giavazzi (1999). In
Germany, Italy and Spain monetary policy operates via a lending channel in
an important segment of the market ± a segment that accounts for more than
80 per cent of the total German banking industry and about half of the Italian
and Spanish industry. The situation is quite different in France, where there is
no evidence of a lending channel.
Cross-country differences in the process of ®nancial intermediation could
be the result of varying national preferences and traditions. Consequently, a
consolidated, cross-border, ®nancial institution may wish to continue offering
different products in different markets. Similarly, the respective roles of
markets and intermediaries may be history-dependent in a way that will not
allow for rapid changes. Nevertheless, the creation of cross-border suppliers of
®nancial services, at a time when European consumers and ®rms are likely to
become more similar, would plausibly result in a homogenisation of ®nancial
practices across EMU.
The current consolidation of the European banking industry appears,
however, to be moving in the opposite direction. It is instructive to observe
what is happening in the light of the parallel experience in the United States.
Despite the massive consolidation which has occurred, concentration at the
local level has, if anything, decreased. Table C2.1 shows the Her®ndahl index of
the concentration of local markets for bank deposits in the United States:
consolidation has been accompanied by no signi®cant change in concentration.
296 Francesco Giavazzi
Table C2.1 Measures of concentration in US banking markets, 1988 and 1997
Per cent of total assets of Her®ndahl concentration indexa

domestic banks held by Metropolitan Statistical Areas Non-MSA
the top eight banks (MSA) counties counties
1988 22.3 2020 4316

1997 35.5 1949 4414
Notes: a The deposit Her®ndahl index is 10 000 times the sum of squared market shares based on
deposits of banks operating in MSA and non-MSA counties
Source: Berger, Demsetz and Strahan (1998), reproduced in Danthine et al. (1999).
Table C2.2 documents the characteristics of the consolidation which has so far
occurred in Europe. Unlike in the United States, most European banking deals
(half of all mergers or acquisitions in 1997) have involved institutions based in
the same country. Cross border activity has been limited to deals involving a
bank and a non-bank ®nancial institution (NBFI), mostly an investment bank,
an insurance company or an asset manager. While cross-border deals are
motivated by the search for experience in corporate ®nance and asset
management ± skills that are in scarce supply in continental European banks
± domestic deals are mostly driven by the search for size.
There are two reasons to be concerned. Competition is the ®rst. European
banks have a natural tendency to consolidate within national boundaries,
leading to industry concentration ratios much above those observed in the
United States. This is because of potential cost-cutting, culture and trust ± and,
indeed, the quest for market power at a time of insecurity and change: sheep
get closer together when in danger. In commercial banking, diversi®cation
gains explain the success of interstate consolidation in the United States. The
anaemia of the equivalent cross-border merger and acquisitions (M&A)
business in Europe is worrisome: while it can be explained by the fact that a
good deal of the gains from diversi®cation can be obtained within the borders
of individual European states, it also generates the concern that European
commercial banks will want to reach the higher minimum size in their
business simply by acquiring or merging with their national competitors. The
Table C2.2 Bank acquisitions in Europe, 1993±7 (value of all deals, US$ billion)
1993 1995 1997
Total 19 40 122
Of which:
± Domestic bank/bank 9 24 60
± Cross-border bank/bank 1 8 7
± Bank/non-bank 9 8 55
Source: Goldman Sachs, reproduced in Danthine et al. (1999).

observed tendency to in-country consolidation is a challenge for competition

authorities as it is likely to reinforce local monopoly power. This is particularly
important for small-®rm lending, as large ®rms will access the Euro capital
markets directly, while consumers will have the option of turning to
specialised asset managers and to direct banking.
Second are the consequences for the monetary transmission mechanism.
The creation of cross-border suppliers of ®nancial services, at a time when
European consumers and ®rms are likely to become more similar, would
plausibly result in a homogenisation of ®nancial practices across EMU. One
would expect, for instance, that the ability of medium-sized Italian ®rms to
make use of the opportunities offered by the emerging Euro-wide ®nancial
markets will be improved if Deutsche Bank were to take over Banca
Commerciale Italiana and the clients of Banca Commerciale could bene®t
from the universal banking experience of Deutsche Bank±Bankers Trust, as
compared with the alternative of a local merger between Banca Commerciale
and Banca Intesa. In the former case, there is reason to believe that traditional
banking relationships will be reconsidered and aligned on the best German
and European practices in a way that would not be promoted by a pure
domestic consolidation.
References
Berger, A. N., R. S. Demsetz and P. E. Strahan (1998) `The Consolidation of the Financial
Services Industry: Causes, Consequences, and Implications for the Future', Board of
Governors of the Federal Reserve System, November, mimeo.
Monetary Transmission Mechanism', Journal of Economic Perspectives, 9, 27±48.
Danthine, J. P., F. Giavazzi, E. L. von Thadden and X. Vives (1999) À Brave New World:
European Banking After EMU', London, CEPR.
Favero, C. A., L. Flabbi and F. Giavazzi (1999) `The Credit Channel and (A)Symmetries in
the Monetary Policy Transmission Mechanism in Europe: Evidence from Banks'
Balance Sheets', Milan, Boconi University, mimeo.
Kashyap A. K. and J. C. Stein (1997) `What do a Million Banks Have to Say about the
Transmission of Monetary Policy?, NBER Working Paper, 6056.
Kashyap A. K., J. C. Stein and D. W. Wilcox (1993) `Monetary Policy and Credit
Conditions: Evidence from the Composition of External Finance', American Economic
Review, 83, 78±98.
Claes Berg1
The connection between a country's legal structure, ®nancial structure and

monetary transmission mechanism made in Chapter 5 in this volume by
Stephen Cecchetti leads to the intriguing conclusion that without legal
harmonisation the ®nancial structure will remain heterogeneous across EMU
member states, and because of this the monetary transmission mechanism
will remain asymmetric within the EMU. Cecchetti has made an attempt at
quantitatively substantiating this conclusion. The evidence can, however, be
viewed only as tentative, as is also clearly stated in Chapter 5; the comments
by Neumann and Angeloni showed that some additional research on re®ning
some of the measures used would be useful.
The crucial issue, of course, is to what extent the asymmetry in the EMU's
transmission mechanism is a problem. I basically agree with what Angeloni
has noted ± namely, that other parts of the transmission mechanism, such as
the interest rate channel and the exchange rate channel, will become more
harmonised in the Euro Area. It is also important to emphasise what Philippe
Moutot said, in commenting on Weber's Chapter 4 in this volume, that the
security settlement system has been harmonised in Europe and also the system
for accepting collateral for the whole Eurosystem. This may have repercussions
on the way the legal system will affect the monetary policy process of the ECB.
However, as Reinhard Schmidt (Chapter 6 in this volume) and Otmar Issing
(p. 283±92) have both pointed out, these countries are no longer small open
economies and their response functions will probably have changed as a
result.
Being a central banker it is encouraging to hear from Bennet McCallum, in
Chapter 1 in this volume, that it is systematic behaviour that matters rather
than unforeseen monetary shocks. McCallum made this point in a simple
three-equation model without optimising behaviour. However, this point can
easily be made in a simple optimising dynamic general equilibrium model,
where the only real effect of monetary policy is through the in¯ation tax. The
effects on welfare of the in¯ation tax can be substantial when the other
distortionary taxes are also modelled. Then adding stochastic shocks to the
monetary policy rule have minor effects on welfare (see Jonsson and Klein,
1997).2
In Michael Burda's Chapter 7, it is claimed that nominal prices will be more
rigid because the sheltered sector of Europe as a whole is relatively larger than
in individual countries where external in¯uences on price-setting have been
important. On the other hand the new European market is much larger in size
than any regional sheltered market, which should improve competition and
¯exibility in pricing in some goods and service markets. In sum, it is dif®cult to
298
give a ®rm conclusion whether ¯exibility of nominal prices will improve or

not in EMU.
The view that institutions are endogenous, and will be affected by the
introduction of EMU, seems reasonable in the sense that more ¯exibility in
real wages could be expected over time. However, as the comments by Brouwer
and Campos e Cunha showed, it seems that the analysis to some extent
underestimated the length of the adjustment process. It is reasonable to
believe that reforms of labour market regulations and other safety nets will
meet political resistance, at least in some countries. The tendency towards less
union power could cause reactions. Furthermore, in the area of ®scal policy
coordination it will take some time before a new equilibrium will be reached.
I will therefore say some words on a somewhat neglected issue ± the possibility
of using 'internal devaluations' as a substitute for monetary policy when
countries are subject to asymmetric shocks in EMU. Variations in payroll taxes
paid by employees could be used to reduce (or increase) production costs without
affecting the overall ®scal balance if other taxes or government expenditures
were changed at the same time. In a recession, relative unit labour costs can be
depreciated through a cut in the payroll tax rate. This could be neutralised (in the
budget) by a rise in employee contributions to the social security system, in
income taxes or in VAT. In Finland, central labour market organisations have
concluded an agreement on so-called 'buffer funds' to help ®nance unemploy-
ment insurance and pensions. The idea is to build up such funds in good times
and draw on them in recessions and thus avoiding raising employer
contributions. The motivation is thus to allow some ¯exibility in wage costs
without having to reduce actual nominal wages (see Calmfors, 1998).
Let me now give some general remarks on monetary policy and the ECB
strategy. In recent years a growing consensus has emerged for price stability as
the overriding, long-run goal for monetary policy. A nominal anchor is a
constraint on the value of domestic money and in some form it is a necessary
element in a successful monetary regime. A public price stability target thus
can serve as a nominal anchor and coordinate in¯ation expectations.3 This
provides the central bank with a direct link to households and ®rms
expectations, which goes beyond the traditional interest rate and exchange
rate channel of monetary policy. An important consequence of the
convergence of in¯ation rates in Europe during the 1990s has probably been
its effect on in¯ation expectations.
In practice, price stability targeting is a package including the following
components: ®rst, announcement of a numerical price stability target which
anchors expectations; second, speci®cation of the target horizon for ful®lment
of the target; third, a clari®cation of the way in which the central bank intends
to meet the target on the relevant horizon; fourth, announcement of any
escape clauses that may warrant departures from the general rule; ®fth,
communication of the strategy to the public using monthly or quarterly
reports, speeches, etc.
300 Claes Berg
When the central bank discusses how it intends to meet the price stability
target there are several different ways to do it. First, the central bank could
publish the in¯ation forecast for the target horizon, given an unchanged
monetary policy stance, and then adjust the interest rate accordingly. This is
done by in¯ation targeting central banks in New Zealand, the United Kingdom
and Sweden. The second and most common way is to regularly publish a
report which discusses price developments, indicators of economic activity
and in¯ation and in¯ation expectations, in a more general way, but not
publish a speci®c in¯ation forecast. This is done by many central banks around
the world, including the ECB.4 In my view, the in¯ation forecast is a crucial
input in monetary policy decisions and will always be very important for a
monetary policy directed at price stability. By publishing the forecast, the
central bank will increase the understanding of monetary policy. It will also
contribute to increased transparency and accountability, as the public at large
will be able to assess whether the stance of monetary policy is appropriate or
not.5
However, the ECB is a new central bank and I guess that it will take time and
experience to organise working routines. I expect that the Eurosystem will
develop its methods in order to be prepared to publish explicit forecasts some
time in the future.6
As was pointed out in Chapter 2 in this volume by Lars Svensson, the
formulation of the monetary policy strategy of the Eurosystem could be
clari®ed. Given the presentation of the reference value for M3 for 1999, it
seems that the implicit price norm is 1±2 per cent. It could be announced that
the ECB aims at a point estimate (for example, 1.5 per cent). In any case, a
symmetric in¯ation target, either a point target or a target range with equal
emphasis on the upper and lower bound, has some obvious advantages in
order to anchor expectations.
Another issue in relation to the ECB strategy ± which can be discussed ± is
the role of gradualism versus activism in interest rate setting. It is widely
accepted that policy makers facing uncertainty about the structure of the
economy should be more cautious when implementing policy. This is the
celebrated conservatism principle due to Brainard (1967). When there is
uncertainty about some structural parameters the policy response usually
becomes more cautious. In Svensson's Chapter 2, this no longer is always true.
When there is non-additive uncertainty and certainty equivalence no longer
holds, some covariance patterns for the parameters may make the optimal
response more sensitive when uncertainty increases. However, Svensson is not
very speci®c, in Chapter 2 at least, for which variables this may be the case in
practice.
One possibility that comes to mind is when there is uncertainty about the
persistence of in¯ation. Then, somewhat paradoxically, it may be optimal for
the central bank to respond more aggressively to shocks than if the parameters
were known with certainty. With persistence in in¯ation, a too small policy
response this quarter will lead to deviations of the target not only next quarter
but also in forthcoming quarters. Therefore it may be optimal to pursue a more
aggressive monetary policy this quarter when the economy is hit with a shock,
in order to avoid bad outcomes in several periods (see So È derstro
È m, 1999).
Another possibility is when policy makers are uncertain about the state of
the economy and learn from the economy's reaction to policy. When the
private sector anticipates systematic attempts to incorporate this information
into policy, modest interest rate changes may prove ineffective. In a recession,
gradual policy initatives may elicit very little reaction. Because small interest
rate cuts are unlikely to end the recession, ®rms and consumers feel safe
waiting for rates to fall again before considering investing/consuming. A
vicious circle may develop in which the expectation that the policy could fail
leads investors/consumers to delay investment/consumption, thereby pro-
moting failure (see Caplin and Leahy, 1996).
This type of reasoning may have some bearing on the situation in part of the
Euro Area, for example in Germany, although nominal interest rates are very
low ± the IFO business expectations and business climate index has been
falling since the end of 1997.
Notes
1. I would like to thank Magnus Jonsson, Yngve Lindh, Peter Sellin and Lars Svensson
for their comments. The views expressed are solely the responsibility of the author
and should not be interpreted as re¯ecting the views of the Executive Board of
Sveriges Riksbank.
2. Regarding the second point, the RBC (or SDGE, stochastic dynamic general
equilibrium) literature have built sound theoretical models and confronted
simulated moments from these models with the data for many years. Even formal
statistical tests have been developed ± for example, SMM. McCallum's Chapter 1 in
this volume gives support for and follows this modelling tradition. Given this, one
question naturally arises (which is not discussed in Chapter 1): why is the central
bank not optimising? There seem to be two dominating strategies for modelling
monetary policy at the moment. On one hand, we have the optimising dynamic
general equilibrium approach where monetary policy usually only follows an (ad
hoc) monetary policy rule. On the other, we have a class of models where the central
bank is optimising but the agents' behaviour is not modelled in any detail. It would
be interesting to see a combination of these two modelling strategies. That is, models
where the central bank is optimising welfare, taking into account its effects on the
agents' behaviour, and at the same time, agents are optimising given the central
banks' behaviour. Of course, this is not an easy task since it involves modelling
dynamic games. But, nevertheless, it would most certainly give new and interesting
insights on how optimal monetary policy should be conducted.
3. A nominal anchor can be implemented in order to avoid both in¯ation and
de¯ation (see Berg and Jonung, 1999) for an account of the early Swedish
experience, 1931±7.
4. Another possibility would be to announce the monetary policy reaction function,
the policy rule or instrument rule, the central bank intends to follow. The rule
302 Claes Berg
should explain how the interest rate should react, given certain deviations of, for
example, actual in¯ation from target, deviations of actual output from potential
output and changes in real exchange rates. However, at least to my knowledge, no
central bank has announced such a reaction function. The reason, I guess, is that
there may be situations when monetary policy will have to deviate from the
announced reaction function.
5. An interesting task for future research is the following: how should the forecast be
aggregated over members when there is an executive board or a governing council?
Voting over forecasts may be very dif®cult. This problem has aspects related to the
problem ®rst discussed by Condorcet in the 18th century. In modern terms, it is
often referred to as the 'jury problem': to decide whether the accused is guilty or not
requires con¯ating the opinion of several experts, with varying competence, into a
single judgement. In practice, central banks ± be them in¯ation targeters or
monetary targeters ± solve this problem. But how is it done?
6. In the Swedish case, in¯ation forecasts have been gradually introduced (see Berg,
1999). The implementation and communication of monetary policy since 1993 can
be divided into three phases. In the ®rst phase, 1993±5, the in¯ation target strategy
was announced and established. However, during this period, in¯ation forecasts
were not published by the central bank; in the reports, the risks for future in¯ation
were stated in a more general way. In the second phase, 1996±7, an in¯ation-forecast
targeting was introduced. The Riksbank's own in¯ation forecasts were given more
weight in the communication of monetary policy. Forecasts for future in¯ation were
gradually introduced. In the third phase, from 1998, 'distribution forecast targeting'
(see Svensson's Chapter 2 in this volume) was introduced and explicit paths for
future in¯ation were published, surrounded by uncertainty intervals.
References
Berg, C. (1999) Ìn¯ation Forecast Targeting: the Swedish Experience', paper presented at
the seminar on 'In¯ation Targeting' in Rio de Janeiro, May, Sveriges Riksbank, Working
Paper.
Berg, C. and L. Jonung (1999) `Pioneering Price Level Targeting: The Swedish Experience
1931±37,' Journal of Monetary Economics, 43 (3), 525±51.
Brainard, William (1967) Ùncertainty and the Effectiveness of Policy', American
Economic Review, 57, Papers and Proceedings, 411±425.
Calmfors, L. (1998) `Macroeconomic Policy, Wage Setting and Employment ± What
Difference Does the EMU Make?' Institute for International Economics Studies,
Stockholm University, Seminar Paper, 657.
Caplin, A. and J. Leahy (1996) `Monetary Policy as a Process of Search', American
Economic Review, 86 (4), 689±702.
Jonsson, M. and P. Klein (1997) `Tax Distortions in Sweden and the United States', in
M. Jonsson, Studies in Business Cycles (IIES, Stockholm University, Monograph Series,
No. 34).
È derstro
So È m, U. (1999) `Monetary Policy with Uncertain Parameters', Sveriges Riksbank,

Working Paper.
Ignazio Visco
1 Introduction
In spring 1998 many commentators would have predicted, in some cases

almost as a fait accompli, that the launch of the Euro would be beset with
problems. It was nonetheless considered that economic prospects for the Euro
Area in 1999 were highly favourable, with predictions of solid growth, low and
steady in¯ation and reduced levels of unemployment. In short, few
imbalances were evident and prospects were rather bright. Indeed, growth
was expected to slow down signi®cantly in the United States, in connection
with the crisis in the emerging economies and the increasing stress in the
®nancial markets. In the fall, while some reduction in growth was also seen as
a possibility in the Euro Area, the overall prospects were still considered to be
rather positive. This led the OECD to recommend in its December 1998
Economic Outlook a cautious easing of monetary conditions in the Euro Area,
and a sharper reduction of policy interest rates in the United States distributed
over 1999. The implications for the Euro were quite obvious, ruling out, under
these conditions, the case for its weakening.
With the bene®t of hindsight, the launch of the Euro was accomplished
without any major hitches, but there have been mounting concerns that
prospects in the Euro Area have deteriorated. If anything, the Euro has been
weaker than expected. The decision by the Federal Reserve to lower interest
rates in three discrete steps of a 14 point between end-September and mid-
November 1998 had the unexpected effect of bringing back substantial
buoyancy in the US stock market. Furthermore, the strength of the US
economy continued to be remarkable in the last quarter of 1998, leading many
to consider a possible tightening of monetary conditions as a new possibility.
At the same time, the substantial deterioration of business con®dence in the
Euro Area and the associated sharp deceleration in current and expected
economic activity revived the calls for further cuts in interest rates by the ECB,
notwithstanding the non-negligible easing associated with the completion of
the process of monetary uni®cation.
Such a characterisation provides an apposite reminder of the hazards of
foretelling even the near-term future. I would add that equally evident are the
large uncertainties which surround the monetary policy transmission
mechanism in the Euro Area and, as a most important consequence, the basis
for clear-cut monetary policy decisions. I will not pretend that I have the
answers to the many challenges which confront monetary policy makers, but
rather raise certain issues and questions which are pertinent to a better
understanding of how monetary policy operates in the Euro Area. My remarks
303
304 Ignazio Visco
will be structured around three themes. First, I shall give a brief summary of
current economic conditions and prospects in the Euro Area. Then I shall
sketch out what I see as some of the key issues and constraints which confront
the design of macroeconomic policy and monetary policy in particular and,
®nally, highlight some of the key questions concerning monetary policy in
the Euro Area.
2 Recent developments and prospects in the Euro Area
The Euro Area experienced a slowdown in growth towards the end of 1998. On
a year-on-year basis GDP growth eased to 2.4 per cent in the ®nal quarter of
1998 (with almost no growth with respect to the third quarter) compared with
about 4 per cent in the ®rst quarter and a solid 3 per cent in the central part of
the year. Leading indicators suggest that activity remained sluggish in the ®rst
half of 1999. The slowdown in activity primarily re¯ected weaker external
demand, associated with recession in many emerging markets and Japan, and
low business investment. While still substantially higher than during the
trough of the 1993 recession, utilised capacity was declining and, while
consumer sentiments remain sustained, business con®dence was rapidly
deteriorating through the second half of 1998. However, weaker growth was
not common to all countries in the Euro Area. In 1998 France recorded its best
GDP growth performance since 1989 and periphery countries such as Ireland
and Finland continued to grow at a pace close to double that for the Euro Area
as a whole.
Divergent in¯ation rates within the Euro Area have also surfaced. In¯ation
measured by Eurostat, the Harmonised Index of Consumer Prices (HICP) is less
than 0.5 per cent in France, Germany and Austria and above 2 per cent in
Ireland and Portugal. The gap between the highest and lowest annual in¯ation
rate in January 1999 rose to 2.3 percentage points,1 compared with 1.4 percentage
points a year earlier. One might ask whether a disparity of this size might pose
a signi®cant problem for setting monetary policy. While some persistent
differences in in¯ation rates are plausible to the extent that they re¯ect a
catch-up process and short-term differentials could be signi®cant, owing to
differences in the relative cyclical position or tax changes, the appraisal would
certainly be more circumspect if the disparity in in¯ation were sustained as a
permanent difference.
Differences in regional price movements within long-established monetary
unions provide a gauge of the potential divergence of in¯ation rates in the
Euro Area. Large in¯ation differentials have seldom been observed in other
monetary unions and when they did happen, they were very short-lived
episodes. For instance, the differential between East and West Germany
following uni®cation was initially almost 10 per cent per annum, but fell
quickly to less than 1 per cent. In the United States, modest differentials of
between 0.7 and 1.3 per cent per annum among state consumer price indices
have been apparent over the last thirty years. Likewise, the average differential
between the lowest and highest in¯ation rate in the major Australian cities
over ten-year periods is well below 1 percentage point; the same holds true
considering the differential across Spanish regions.
While in¯ation differentials in economically homogeneous areas are
unlikely to accumulate to large price level differences, in¯ation differentials
could persist between the more and less advanced Euro Area economies. As
is well known, productivity differentials between traded and non-traded
goods might not only generate sectoral in¯ation differentials, but differ-
ences in in¯ation rates between countries at different stages of develop-
ment. One might consider this to be a possibility even for the Euro Area:
indeed GDP per capita ranges between 12 000 Euros in Portugal and 29 700
Euros in Luxembourg (17 400 Euros for the area as a whole), even if the
variation is much less if one adjusts for PPP differences. Other indicators of
living standards ± such as the number of doctors, passenger cars, telephones
and television sets per 1000 inhabitants ± also vary to a non-negligible
extent.
While some evidence of divergent economic conditions and disparate levels
of development is apparent across the Euro Area, much progress towards
structural and cyclical convergence has been achieved. Indeed, among the
largest three economies in the Euro Area ± Germany, France and Italy ± which
account for some three-quarters of total output, the level of economic
development and living standards are very similar. The Maastricht criteria also
suggest that closer convergence has been obtained over more recent years.
Actually one of the most striking developments has been the rapid pace at
which Euro Area countries reduced their ®scal de®cits and accomplished the
nominal convergence criteria (in¯ation and interest rates) required by 1997
(the reference year). This achievement was con®rmed in 1998. Indeed, in
spring 1999 only Ireland, the Netherlands and Portugal (accounting for less
than 10 per cent of the Euro Area GDP) have in¯ation rates slightly higher
than 2 per cent. With Finland, they are the only Euro Area countries with a
positive and growing output gap.
In general terms, a monetary union is less susceptible to asymmetric
disturbances the more its regions are integrated with each other and
diversi®ed within themselves. Apart from the Maastricht convergence
criteria prior to the launch of the Euro, there are three key forces that
promote closer integration. These are the degree of trade interdependence,
the degree of intra-industry trade and closer income linkages, such as
increased foreign direct investment or closer ®nancial market interactions.
Over the past twenty years each has risen sharply and promoted closer
synchronisation of business cycles across Euro Area countries.2 This process
of integration is expected to continue in the years to come, and is possibly
accelerated by the introduction of the Euro.
306 Ignazio Visco
3 Some monetary policy challenges and issues
Although disparities at the periphery of the monetary area raise a number of

issues, the main problem in the current conjuncture is the risk of a signi®cant
slowdown in the area at large. The combination of slower growth in economic
activity and weaker prospects, falling in¯ation and the constraints on
expansionary ®scal policy to absorb slack in the Euro economy have led to
calls for an easier stance of monetary policy. Most importantly, it is necessary
to raise the Euro economy's adaptability, and reduce its susceptibility to
shocks and policies with uneven impacts across the monetary area. This will
require further progress in implementing structural reforms, particularly
enhancing the ¯exibility of labour markets.
The calls for a softer monetary policy stance are (obviously) not uniformly
accepted. The main arguments against a reduction in the policy interest rate
include actual monetary conditions, which some already consider expan-
sionary. Many Euro Area countries experienced fairly sharp declines as short-
term interest rates converged in the second half of 1998, with falls amounting
to a 75 basis-points reduction for the area, on average. Yields at the longer end
of the maturity spectrum also fell sharply, to around 4 per cent in 1999
compared with 512 per cent a year earlier. Overall, real interest rates, both short
and long, were substantially lower than in the ®rst half of the 1990s (between
2 and 3 per cent now, compared with levels between 6 and 7, if not more, in
the early 1990s). Furthermore the Euro real (and nominal) effective exchange
rate depreciated by just over 5 per cent between October 1998 and March
1999. Moreover, monetary and credit aggregates are growing solidly. The
lagged effects of these conditions should lend some support to domestic
demand. Based on OECD simulations of the Interlink model, the 5 per cent
depreciation of the Euro could raise the level of GDP by about 12 a per cent and
only marginally boost in¯ation after two years.
It should be borne in mind, however, that the level of the Euro in real
effective terms is little changed from its level at the beginning of 1998 until
spring 1999. Furthermore, other indicators of the overall stance of monetary
policy suggest some scope for easing. The output gap in the Euro Area has not
changed much since 1995 and is unlikely to do so in 2000±2001. Indeed, the
low and fragile levels of business con®dence could signi®cantly tilt growth
prospects downwards and result in greater spare capacity and further
downward pressure on output prices. Moreover, prospects beyond the Euro
Area are expected, on the whole, to remain relatively sluggish, which should
limit external in¯ationary pressures. A simple relation between EU-wide
in¯ation, the output gap and import price in¯ation suggests that pressures on
prices are currently more on the downside than the upside.
But how much faith can we place on all these indicators of economic and
monetary conditions? We are inexperienced in analysing and interpreting
Euro Area data. In particular, past observed relations may have changed (or are
changing) with the launch of the Euro. Moreover, few historical series exist on
which to base fairly conclusive analysis. To give an example, how accurate is
the claim that real interest rates are near historical low levels in the Euro Area?
This appears to be the case when real rates are measured based on past
movements in the consumer price index (CPI), but quite a different picture
emerges when alternative de¯ators or in¯ation expectation measures are used.
For example, the short-term (long-term) real rate for the Euro Area using
headline HICP in¯ation is 2.3 (3.2) per cent. But, if a business investment
de¯ator were adopted, the measured real rate might be up to a full percentage
point higher (and even more, if producer prices were used). On the other hand,
excluding commodity and energy prices from headline in¯ation might lead to
measures of short-term real interest rates below 2 per cent.
In addition, it is well known that all CPIs have some inescapable positive
bias. In the United States estimates of the bias range between 0.5 and 1.0 per
cent per annum. The design and compilation of the HICP index, used by the
European Central Bank (ECB) to monitor in¯ation, attempts to limit the
source of known biases, but it is still believed to be positive. This implies that
the level of in¯ation at 0.8 per cent may in fact be closer to the lower end of the
range implicit in the ECB's interpretation of price stability, and this would
raise important questions concerning the stringency and symmetry of the
ECB's monetary policy objective.
While monetary policy in a single currency area, by de®nition, can respond
only to area-wide in¯ation pressures, sovereign ®scal policy can respond to
limit the contractionary or expansionary impacts of regional shocks. This
response is, however, limited by the provisions of the Stability and Growth
Pact. Based on a variety of techniques, further progress on budget consolida-
tion is required in order to avoid èxcessive' de®cits and to provide adequate
scope for ®scal stabilisers to fully operate in the event of a slowdown in
economic activity. Most estimates suggest that a ®scal position on average
close to balance would be necessary to this end. This compares with structural
general government ®nancial balances that still exceed, for the average of the
Euro Area, 1.5 per cent of GDP.
Fiscal policy plans over the ensuing two to three years are targeting declines
in de®cits (from 2.3 per cent of GDP in 1998, to about 1 per cent of GDP). This
is in line with the principles of the Stability and Growth Pact. Furthermore,
most countries are still near or exceed the limit of the general government debt
to GDP ratio of 60 per cent speci®ed in the Maastricht Treaty. Although the
®scal plans embody tight controls on public spending, they are also based on
somewhat optimistic growth projections and in some cases `backloaded'. A key
risk to their realisation is a larger than expected and prolonged slowdown in
activity, which constrains automatic stabilisers from operating without
breaching the Pact's 3 per cent de®cit to GDP limit. Such a scenario, however,
would imply substantially lower GDP growth. OECD simulations suggest that
a 1 percentage-point reduction in output growth would result in an
308 Ignazio Visco
approximately 12 percentage point increase in the de®cit to GDP ratio. On this

basis, GDP growth in the Euro Area would need to be almost 2 percentage
points lower than the projections contained in the December 1998 OECD
Economic Outlook. Scenarios of this kind still appear somewhat unlikely. But
some Euro Area countries, notably the three large economies, have de®cit
levels in 1998 closer to 3 per cent, and thus there is clearly little room for
discretionary ®scal policy. And the risk that relatively depressed levels of
economic activity may persist for some time, in the face of further declining
levels of business con®dence and insuf®cient ®xed investment, calls for
attention on the part of policy makers.
Ultimately the issue consists in how well policies will respond to limit these
downside risks. Recourse to monetary policy, obviously, cannot be expected to
serve as a panacea for all problems ± and, in particular, asymmetric
disturbances. Looking at recent experience, the reduction in con®dence
(and associated low investment levels and growth prospects) cannot be
attributed to a restrictive monetary policy stance. Given that use of ®scal
policy is rather limited, structural reforms need to be pursued so that
adjustment can take place without putting undue strain on economic activity.
It is in particular important to promote greater overall labour market ¯exibility
that would support job mobility and accelerate the pace at which real wage
adjustments take place. Euro Area countries will undoubtedly continue to
experience economic disturbances that affect a single country or sub-sets of
regions. Such country- or region-speci®c supply and demand disturbances
could be sources of tension within EMU unless they can be easily absorbed. It
is in all parties interest to avoid such tensions. One approach towards this end
is to strengthen the coordination of policies (structural, monetary and ®scal).
A rapid and transparent discussion between the monetary authorities and the
Euro Area governments on the appropriate policy mix ± inclusive of the
structural reform requirements ± while preserving the full independence of
the ECB would be a valuable step forward.
4 Conclusions
These remarks highlight the many uncertainties and much uncharted water
ahead concerning the framing of monetary policy in the Euro Area. To
complicate matters further we are still in a learning process about the
transmission channels and the effectiveness of monetary policy in the new
monetary area. Answers to, or at least a better and deeper understanding of,
the following questions might help reduce some of these uncertainties:
1. How effective would a small interest rate cut be in the current context? In
particular, what can be learned from US experience ± that is, what can be
said about the effects on economic activity stemming from con®dence,
expectations and asset price movements?
2. Is the transmission mechanism different when sizeable slack exists in an

economy? Or, more speci®cally, is there a range over which the use of
monetary policy has no direct effect on prices, so that interest rates can be
reduced without in¯ationary consequences and with positive effects on the
real economy?
3. How does one interpret monetary conditions and what are the implications
for setting policy?
With respect to the ®rst point, what would seem the most important is the
need to avoid possible cuts in interest rates ending up reducing rather than
improving con®dence. This calls for improved policy coordination efforts
which would also be helpful on the communication side: this would prevent
the perception that the easing of monetary policy would either re¯ect the fact
that the ECB had given up on the expectation that governments were ready to
act rapidly and effectively on the structural front, or signal a state of current
and prospective economic conditions much worse than expected.
The second point is, perhaps, even more dif®cult to deal with. This is simply
because there is still not much information on how the Euro Area economy
operates ± i.e. how economic decisions might differ under a common
monetary policy compared with the previous state of affairs. What might be
relevant here is probably the need to avoid the ECB's concerns about price
stability being perceived as biased towards less favourable monetary condi-
tions than would be needed in a situation of substantial current and
prospective slack in the Euro Area as a whole. To this end, the (perhaps
evolving) identi®cation of a lower bound for the de®nition of price stability
might be helpful (at least operationally). Moreover, it is important that the
ECB should not be perceived as feeling equally at ease with price changes
permanently situated at the bottom and at the centre of the range.
Finally, while many measurement problems must still be sorted out, both
with respect to price indices and monetary aggregates, a wide spectrum of
interest rates should be considered. In fact, one can always expect to ®nd
con¯icting signals concerning the stance of monetary conditions, depending
on the indicators used. In addition the Euro Area has virtually no historical
perspective against which benchmarks can be established. Consideration
should then also be ± and undoubtedly will be ± given not only to the levels
and spreads prevailing in the Euro Area member countries in the course of the
1990s, but also to those prevailing in other economies at different points of
the cycle, as well as over a longer time perspective.
Notes
1. Portugal has the highest in¯ation rate at 2.5 per cent and Luxembourg the lowest at
±1.4 per cent. The reported differential, however, excludes Luxembourg since the
rate in January has been in¯uenced by a one-off effect almost entirely owing to the
inclusion of the January discount sales in the price index for the ®rst time.
310 Ignazio Visco
2. The OECD has just published an in-depth assessment of the challenges facing the
European Single Currency Area, including an evaluation of the forces promoting
closer economic integration (OECD, 1999).
Reference
OECD (1999) EMU: Facts, Challenges and Policies, Paris, OECD.
Manfred J. M. Neumann
I will brie¯y take up two issues: (1) What type of ®nancial sector structure is
preferable? (2) Should the European Central Bank (ECB) be concerned about
differences in regional or national transmission mechanisms?
1 What type of ®nancial sector structure is preferable?
Structure of the ®nancial sector

A central theme of this conference has been that the effectiveness of monetary
policy, as regards impact as well as speed of transmission to the real economy,
depends on the structure of the economies' ®nancial sectors. Though the
available empirical evidence is not overwhelming, most economists ± and
certainly the supporters of the `credit view' ± share the opinion that monetary
policy is transmitted to the real sector quicker and more strongly if an
economy's ®nancing relations are dominated by a large and not too healthy
banking sector. Thus, it seems that, from the monetary control point of view,
bank dominance is the preferable state.
While this is a probably important message for central banks, provided they
desire to contribute to cyclical stabilisation, it is not obvious that bank
dominance is equally preferable from the long-run view of economic
development. The long-run success of an economy as regards real growth
and the creation of jobs depends, apart from other factors, on the ¯ow of
innovation. From this more long-term point of view one might question the
value of bank dominance. As a rule, banks are relatively risk averse institutions,
as they have to be in the interest of depositors. As a result, banks are not too
keen on ®nancing young, innovative companies. The only asset of a young
enterprise in most cases is just a new business idea. Given the lack of assets that
could be used as collateral for bank loans, the lack of pro®ts during the ®rst
years of operation, the extremely high uncertainty about the success of a new
company's attempt at establishing a market for the product innovation, young
companies ®nd it dif®cult to attract the ®nancial means for rapid expansion if
the ®nancial markets are dominated by commercial banks. As a rule, in such
economies venture capital markets are underdeveloped. Thus, from the point
of view that the ¯ow of innovation is an important factor for strong long-term
development, it seems that a preferable ®nancial sector structure is one where
large, liquid markets for the participation in venture capital and equity ®nance
exist.
It is instructive to compare on this account Germany with the United States.
In the United States, it is easier for young companies to attract capital given
that there equity and venture capital markets play a much more important
311
role. This is highlighted by the following facts: (1) in Germany it takes 42 years,
on average, until a ®rm is admitted to the share market by a regulatory
committee that is dominated by banks; in the United States, in contrast, it takes
14 years on average; (2) in the United States market capitalisation, as a
percentage of GDP, is three times of what it is in Germany; (3) the share of bank
lending in total ®nance is much smaller in the United States than in Germany.
In line with these differences in ®nancial structure are the differences in legal
protection. In the United States the private investor in equity ®nance is better
protected by more extensive rights for controlling company management than
in Germany. As a corollary, creditors enjoy better protection under German
than under US law. How have the ®nancial cultures come to develop so
differently? A very probable explanation is the different role of universal banks.
Since the 1870s universal banks have dominated the German culture of ®nance
and this has in¯uenced German law while in the United States the banking
reform of the early 1930s put an end to the existence of universal banks and
promoted the fragmentation of the banking system.
But note that we lack systematic knowledge about the relative advantages as
regards the long-run impact of alternative structures of the ®nancing system.
All in all, it seems that bank-dominated Germany has not fared worse than the
more equity ®nance-oriented United States. While the long-term development
of economies obviously does not solely depend on the quality and structure of
the ®nancial system but on a host of other factors, too, it is interesting to note
that the German economy has not grown less, but faster than the Anglo±
Saxon economies. On average over the past ®fty years, real per capita GDP has
grown by 3.5 per cent in Germany but by about 2 percent in the United States
and in the United Kingdom.
Impact of EMU on the convergence of national transmission mechanisms

The launch of the Euro will speed up the process of ®nancial market
integration in Europe and this, in turn, will promote the convergence of the
transmission mechanisms. As the Euro promotes the establishment of
uniform market standards in the Euro Area, the national ®nancial markets
will rapidly integrate into a larger market that will be able to provide a wider
maturity range of instruments and an increase in market liquidity. This, in
turn, is likely to induce domestic and foreign institutions to expand the
issuance and the investment in Euro instruments. A Euro-wide market for
commercial paper is emerging and the issuance of corporate bonds is
expanding. Similarly the fragmentation of equity trade is vanishing and it is
very likely that the national regulation of equity issuance will be harmonised.
The rising integration and size of the European ®nancial markets provides
bene®ts of scale, and the increase in the degree of competition will reduce
transaction costs.
The national banking industries are involved in these developments and
they are likely to demand that national governments harmonise banking
regulation. Banks will become more similar across the Euro Area and move
into new areas of ®nance. As the Banking Report of the ECB (ECB 1999) shows,
banking intermediation is losing out to institutional investors, like investment
funds, insurance companies and pension funds. Thus, I conclude that the
transmission mechanism in the Euro Area will indeed become more similar to
that of the United States. We can therefore expect that monetary policy will
lose its strength to some degree in the longer run.
2 Should the ECB be concerned about differences in regional or

national transmission mechanisms?
The Treaty of Maastricht assigns two objectives to monetary policy. The ®rst is
maintaining price stability. Provided that this objective is not endangered, the
European System of Central Banks (ESCB) is expected to support the general
economic policies in the Community. Though this second stipulation is open
to interpretation, it probably aims at business cycle stabilisation. The ECB puts
emphasis on the objective of maintaining price stability by calling it the
overriding objective.
With respect to this objective differences in regional or national transmis-
sion mechanisms are of no particular relevance. Consider that EMU starts from
a state of price stability and let us assume that the ECB has adopted a medium-
run strategy of supplying a stable monetary expansion path ± relative to the
evolution of money demand. Under those equilibrium conditions all member
countries are affected to the same degree. Asymmetries in the transmission
mechanisms will not matter, given that the ECB's forecast errors as regards the
appropriate medium-run path are likely to be comparatively small. But note
that we cannot be sure that the ECB has adopted a medium-run strategy. The
ECB has announced that it will take into consideration its reference value for
the growth of the money stock M3 and at the same time a host of other
economic indicators.
In contrast, differences in the transmission mechanism will matter if the
ECB produces monetary shocks which by de®nition are unanticipated. This
will become particularly relevant should the ECB try to ®ght a business
downswing or upswing. The regional economies will be hit differently,
implying that their relative positions as regards competitiveness in the single
market will be affected. This applies especially to small and medium-sized
®rms who traditionally rely on bank ®nance. Moreover, it seems that those
member economies where the bank lending channel is strongest will be hit the
harder if their business cycles run counter to the average cycle showing up in
the EU-11-wide aggregates. This will be a problem for the smaller members of
EMU whose data enter the aggregate data with small weight.
The question is: what lesson should the ECB draw from this perspective? It
goes without saying that national business cycles are none of its business.
Monetary policy has to be based on the EU-11 aggregates and nothing else.
However, the ECB should not completely disregard differences between

national cycles because the larger the differences become the more asymmetric
will be the impact of its policy. In view of this I conclude that the existence of
differences in the transmission mechanism lends additional support to the
monetarist prescription that the ECB should avoid larger policy swings but try
to run a stable monetary policy tied to the medium run.
Reference
ECB (1999) Possible Effect of EMU on the EU Banking System in the Medium to Long Term,
Frankfurt/Main.
Name Index
Note: page numbers in bold refer to main contributions in this book; page
references to ®gures and tables are given in italics.
Agell, J. 266 Cukierman, A. 265

Akerlof, G. A. 67, 93±4 n14, 115, 259 Currie, D. 72
Andersen, T. M. 28
Angeloni, I. 201±7, 298 Danthine, J.-P. 172, 265
de Bondt, G. J. 181
Ball, L. 65, 66 DeFina, R. 285
Bank for International Settlements 1, 206, Dickens, W. T. 67, 93±4 n14, 115
249 Dohse, D. 257, 267
Barro, R. 265 Dornbusch, R. 172, 181, 209, 257
Batini, N. 77 Dotsey, M. 38
Bean, C. 266 Drif®ll, J. 264, 265
Benhabib, J. 115
Berg, C. 8, 298±302
Ehrmann, M. 183
Bernanke, B. S. 11, 31, 33, 38, 78, 132,
Eichenbaum, M. 28, 31, 33, 47, 175
175
Eichengreen, B. 257
Blanchard, O. 256
Eliasson, A.-C. 27
Blinder, A. S. 51, 72, 175
Evans, C. L. 28, 31, 33, 175
Brainard, W. 85, 300
Brouwer, H. J. 7, 276±8, 299
Brunner, K. 115 Favero, C. 159±63, 172, 181, 209, 257,
Bruno, M. 255 295
Burda, M. C. 7, 252±73, 276±8, 279±82, Feldstein, M. 257
298 Fisher, M. E. 140
Buti, M. 257 Flabbi, L. 295
Friedman, M. 255, 283, 291
Calmfors, L. 260, 264, 265 Fuhrer, J. C. 24, 27, 28, 31, 54
Campos e Cunha, L. 7±8, 279±82,
299 Gali, J. 12, 48, 131, 132, 135, 137, 137t,
Carlino, G. 285 140
Cecchetti, S. G. 5, 63, 170±93, 195±200, Gerlack, S. 181
201±7, 242, 298 Gertler, M. 11, 12, 38, 48, 131, 132, 135,
Chari, V. V. 28, 47 137, 137t, 140, 175
Christiano, L. J. 2, 28, 31, 33, 37, 44±9, Ghironi, F. 55
174±5 Giavazzi, F. 8, 172, 181, 209, 257,
Clarida, R. 12, 48, 131, 132, 135, 137, 294±7
137t, 140 Goodfriend, M. S. 31, 168
Cooley, T. F. 45 Goodhart, C. A. E. 6, 241±6
Corbett, J. 222, 223 Gordon, D. 265

Index compiled by Sue Lightfoot.
315
316 Name Index
Greenspan, A. 73, 111 Meltzer, A. H. 3±4, 11, 112±29

Gust, C. 48±9 Mihov, I. 31, 33
Mishkin, F. S. 11, 78, 139
Hackethal, A. 222, 241 Monticelli, C. 289
Haldane, A. G. 77 Moore, G. R. 24, 27, 28, 31
Hall, R. E. 28, 64 Moutot, P. 4, 164±9, 298
Hansen, G. D. 45 Muellbauer, J. 285
Hansen, L. 47 Mundell, R. 252, 253, 279
Herrmann, H. 1±9
HM Treasury 66 Nelson, E. 13, 14, 16±17, 24, 28, 34
Hunt, J. 265 Neumann, M. J. M. 5±6, 8±9, 195±200,
298, 311±14
Isard, P. 27
Issing, O. 8, 89, 90, 91, 283±92, 298 Obstfeld, M. 11
Organization for Economic Cooperation
and Development 303, 307±8
Jeanne, O. 267
Orphanides, A. 67, 94 n14, 115, 153
Jenkinson, T. 222, 223
Peersman, G. 289
Kashyap, A. K. 170, 176, 204 Perez Quiros, G. 183
Kehoe, P. J. 28 Perry, G. L. 67, 93±4 n14, 115
Kieler, M. 181, 288 Posen, A. S. 137, 138, 138t
King, M. 3, 103±6 Prescott, E. C. 47, 51
King, M. A. 65, 88, 94 n26
King, R. G. 47, 132, 139, 140, 142, 143, Rich, R. W. 183
144, 154±5 Rogoff, K. 11
Koenig, E. F. 125±6 Romer, C. 36±7
Krieger-Boden, C. 257, 267 Romer, D. 36±7, 256, 259
Kuttner, K. N. 137, 138, 138t Rotemberg, J. J. 13, 46, 175
Kydland, F. E. 47, 51 Rudebusch, G. 77, 137
La Porta, R. et al. 172, 185, 186, 187 Saarenheimo, T. 181, 288

Lafontaine, O. 270 Sachs, J. 255
Laxton, D. 27 Sapir, A. 257
Leamer, E. 166 Sargent, T. 47
Leeper, E. M. 37 Schmidt, R. H. 6, 208±36, 241±3, 246,
Levin, A. 72 247±50, 298
Levine, P. 72 Schmitt-Grohe, S. 115
Lippi, F. 265 Schuberth, H. 183
Lothian, J. R. 139 Seater, J. J. 140
Shapiro, M. D. 37
McCallum, B. T. 2, 11±40, 44±5, 46±9, Shiller, R. J. 164, 169n2
52, 53, 54±8, 72, 122, 138, 143, 298 Sims, C. A. 37
McCauley, R. N. 172 Smets, F. 181, 289
McConnell, M. M. 183 Smets, G. 249
McGrattan, E. R. 28 Stein, J. C. 170, 176, 204
McKinnon, R. 252 Stephens, M. 285
Maclennan, D. J. 285 Summers, L. H. 115, 139
Mankiw, N. G. 259 Svensson, L. 2±3, 60±97, 103±6, 113,
Mayer, C. 221±2 137, 160, 300
Name Index 317
Taylor, J. B. 11, 28, 57

Watson, M. W. 38, 47, 132, 139, 140,
Tobin, J. 211
142, 143, 144, 154±5
Tristani, O. 289
Weber, A. A. 4, 131±56, 161,
164±8
Uhlig, H. 2, 51±9
White, W. R. 172
Uribe, M. 115
Wieland, V. 67, 72, 94 n14, 115
Williams, J. C. 72
Vickers, J. 78
Vienney, A. 6±7, 247±51

Williamson, O. E. 218
Ä als, J. 3, 107±11
Vin
Wolman, A. L. 94 n14
Visco, I. 8, 303±10
Woodford, M. 13, 46, 80, 132, 168
Wyplosz, C. 151
Vlaar, P. J. G. 183
Wallis, K. F. 78, 79, 96 n38
Walsh, C. E. 27
Yellen, J. L. 259

Subject Index
Note: page references to ®gures and tables are given in italics, e.g. 20f, 32t.
aggregate demand and supply 113, 124 channels of monetary policy 210±11
asset price channel see channel of channel of relative prices 211±12
relative prices credit channel see credit channel
exchange rate channel 211
Bank of Canada 64, 77 interest rate channel 211, 212, 248±9
Bank of England collective bargaining 264±5, 267, 277,
forecast targeting 76, 78±9, 82, 86, 92, 280
106 Consumer Price Index 66±7
in¯ation targeting 66, 67, 86, 95±6 consumption 125±7, 126±7f
n37, 96 n38, 105, 106 credit channel 125, 204±6, 212±15,
bank reserves 174±5 228±9, 247±8, 249, 294
banking industry balance sheet channel 213, 214, 228
change 206, 290±1 bank lending channel 176, 213±14,
consolidation 294, 295±7, 296t, 312±13 228, 295
cross-country differences 171±2, 180t, cross-border competition 205
214, 295 France 249, 295
France 291, 295 Germany 227±8, 228f, 241±2, 243f,
Germany 219, 220, 222, 295, 312 247, 249, 295
(see also Deutsche Bundesbank) Italy 295
health 178, 179, 179t Spain 295
Italy 295 strength 178, 181t, 186, 187, 196,
lending channel 176, 213±14, 228, 197, 198±200, 199f
295 United Kingdom 227±8, 228f, 241,
liability diversi®cation 204±5 243f, 247, 249
size and concentration 177±8, 177t,
181t, 196±7, 197t, 198
de¯ation 68, 115±24
Spain 295
Deutsche Bundesbank 61, 106, 135,
United Kingdom 219, 222 (see also
146, 153
Bank of England)
United States 295, 296t, 312 (see also
Federal Reserve System (USA)) ECB see European System of Central
see also Bank of Canada; credit channel; Banks
European System of Central EMU see European Monetary Union
Banks (ESCB); New Zealand, ESCB see European System of Central
Reserve Bank of; Sveriges Banks
Riksbank European Monetary Institute 67, 73
European Monetary Union (EMU) 195,
capital mobility 254, 254t 210, 303, 306
channel of relative prices 211±12 economic development 305

Index compiled by Sue Lightfoot.
319
320 Subject Index
European Monetary Union (continued) differences 215±19, 216f
effects on labour market 252±3, ®nancing patterns 217, 221±3, 222t,
258±71, 276±8, 280±2

224f, 233t
®scal policy 307±8

France 220, 222t, 223, 224f, 241, 247,
growth 304
250
in¯ation rates 304±5

Germany 219, 220, 222±3, 222t, 224f,
integration 305
225±9, 228f, 241, 243f
interest rates 306, 307

and legal structures 172±4, 173t, 185,
monetary policy 306±9

186, 187t, 188, 191 n5, 202
and national ®nancial systems 223±31,

national differences 201±2, 209,
249±50
214±15, 285
and optimal currency areas 279

non-bank ®nance 178, 179, 180t, 197,
transmission mechanisms 284±8,

198t, 214
286t reactions to common monetary
European System of Central Banks (ESCB)

policy 188, 223±31, 249±50
challenges 170, 188, 230±1

sub-systems 216±18, 216f, 221±2, 247
credit channel 206

United Kingdom 219±20, 222±3, 222t,
de®ning price stability 67, 88±9,

224f, 225±9, 228f, 241, 243f, 250
104±5
see also banking industry
indicators of risks to price stability 81,

Fisher effect 139±40, 166
82
forecast targeting see targeting
in¯ation 124, 259

France
M3 de®nition 205
banking industry 291, 295
maintaining price stability 89±92,

corporate sector 285
290
credit channel 249, 295
money 89±90, 105

®nancial systems 220, 222t, 223, 224f,
monetary policy 195, 282, 289t

241, 247, 250
price stability 66, 88, 89, 104, 105,

interest rate channel 248
313
strategy 283±4, 290, 291±2, 299±300,
Germany
313±14
banking industry 219, 220, 222, 295,
transparency 92
312 (see also Deutsche
exchange rates 202±4, 255, 256, 259f,
Bundesbank)
279, 291, 306
credit channel 227±8, 228f, 241±2,

exports 204t
243f, 247, 249, 295
external ®nance premium 213
®nancial sector structure 311±12
®nancial systems 219, 220, 222±3,
Federal Reserve System (USA) 58±9,

222t, 224f, 225±9, 228f, 241, 243f
115±16, 122, 124, 285

interest rate channel 227±8, 228f,
history 116±24
241±2, 243±5f, 246, 246t, 248
®nancial markets 290±1, 312

goods markets 284±5, 286t
see also banking industry
®nancial sectors 209, 214±15, 231t
national differences 285, 286t, 287,

Haberler±Pigou±Patinkin wealth
311±12
effect 124, 127
®nancial systems 209±10

housing sector 285
complementarity and
consistency 217, 218, 221f,

in¯ation 48±9, 63, 65, 132, 134f,
222, 241
259±61, 260t, 261t, 278, 304±5
corporate governance 232t costs 124
Subject Index 321
and interest rates 133±4f, 136f,

macroeconomic impact 253±8
138±48, 141±2t
structure 286t, 287
econometric framework 143±4, 159,

labour mobility 253±4, 270, 280
161±3, 165±6
labour unions 264±5, 267, 277, 280
empirical evidence 144±8, 145t, 147f,

legal systems
148f, 149f
and impact of monetary policy 173,

173t, 186±8, 187t, 198±200, 205±6
time series properties 140±2

national differences 183, 185±6, 185t,
in¯ation targeting see monetary policy 188, 202
interest rate channel 211, 212, 248±9

see also ®nancial systems and legal
France 248
structures
Germany 227±8, 228f, 241±2, 243±5f,
lending channel see credit channel
246, 246t, 248
liquidity traps 113±15
United Kingdom 227±8, 228f, 242,
243±5f, 246, 246t, 248

Marshall±Hicks rule 264, 272 n15
interest rate targeting see monetary

models of monetary policy 54, 298
policy
diagnostics 31±6, 32±3f, 32t, 34f, 35f,
interest rates
36f, 54±5
Euro Area 306, 307

limited participation model 48
European links
maximum likelihood 47
empirical evidence 151±3, 152t,

`price puzzle' 55±6
167
second moments 47
time series properties of

simple analytical 13±15
differentials 148±9, 167

speci®cation 19, 23±31, 23t, 25±6f,
VAR model 149±51, 161±2, 164±5

27t, 29±30f, 54, 55
exchange rate effects 202±4

structural 15±19, 18t, 20±2f, 48±9,
and housing sector 285

53±5, 56±7f
and in¯ation 133±4f, 136f, 138±48,

testing 47±9
141±2t VAR systems 14±15, 33, 52±3, 54,
econometric framework 143±4, 249±50

159, 161±3, 165±6 monetary policy
empirical evidence 144±8, 145t, communication 62, 92, 108, 109, 111
147f, 148f, 149f

decision framework 61±2, 72±3, 92,

103±4, 106, 108±9, 110±11
time series properties 140±2

economic theory 53, 58
policy 131±5, 168±9, 242 (see also

effects 15±19, 56±8, 170±1, 195,
monetary policy)
208±9
smoothing rules 135±8, 137t, 138t,
Euro Area 306±9
142, 146, 154
impact on output and prices 173t,
wealth effect 124±5
181±3, 182f, 184t, 187t, 198±200,
intermediate targeting see monetary

199f, 206
policy
in¯ation targeting 61, 63±6, 67±8, 88,
internal devaluations 299

103, 104, 105, 115, 300±1
investment 53±4
intermediate targeting 61, 86±8, 103,
Italy 295
104, 110
interest rate targeting 69±73
labour markets
legal systems and 173, 173t, 186±8,
Euro effects on 258±71, 276±8,

187t, 198±200, 201±2, 205±6
280±2
monetary targeting 61
institutions 265, 278, 281, 287, 299

price-level targeting 63±5, 68, 104
322 Subject Index
monetary policy (continued) single monetary policy 188, 223±31,

rules 51±2, 56±8, 62, 71±3, 103±4, 249±50, 290
109±10, 168±9 Spain 295
tightening 288, 294 Stability and Growth Pact 307
see also channels of monetary policy; Sveriges Riksbank
®nancial systems; forecast forecast targeting 76, 78, 79, 82, 86,
targeting; models of monetary 92, 96 n38
policy; shocks to monetary in¯ation targeting 66
policy; single monetary policy; Sweden 64
transmission mechanisms see also Sveriges Riksbank
monetary unions 304±5
money 89±90, 105, 294 targeting
money growth indicator 82, 90, 95 n32, forecast targeting 60±1, 73±9, 109±11;
136f distribution forecast targeting
85±6, 104; judgemental
New Zealand, Reserve Bank of 67, 77, 92 adjustments 77±8; mean,
nominal frictions 255±6 median or mode 78±9, 105±6;
non-additive uncertainty 84±5;
optimal currency areas 279 non-linearities 82±4;
optimality criterion 75±6
price rigidity 256, 256f, 258±9, 259f, in¯ation targeting 61, 63±6, 67±8, 88,
266, 281 103, 104, 105, 115, 300±1
price stability interest rate targeting 69±73
de®ning 60, 62±8, 88±9, 104±5, 107 intermediate targeting 61, 86±8, 103,
indices and levels 66±8 104, 110
instrument assumptions 76±7 monetary targeting 61
loss function 62, 65±6, 68, 69, 71, 73, price-level targeting 63±5, 68, 104
78 taxation 266, 299
versus low in¯ation 62±5 transmission mechanisms 11, 224
maintaining 60, 68±9, 89±92, 107±8, asymmetry 188, 298, 313
109, 290 channel of relative prices 211±12
price-level targeting 63±5, 68, 104 channels classi®cation 210±15
risks 81±2, 90 consumption 125±7, 126±7f
role of indicators 79±82 credit channel see credit channel
targeting 299±300 econometric evidence 124±8
price-level targeting see monetary policy effects of single currency 288, 290±1
product market integration 255, 265 in Euro Area countries 284±8, 286t
exchange rate effects 202±4
rigidities ®nancial structures 170±4, 173t, 176
nominal 255±7, 256f, 258±61, 266, (see also ®nancial systems)
267, 281 historical evidence 115±24
real 253±7, 256f, 262, 264±5, 267 institutional features 284±8, 286t
interest rate channel 125, 211, 248±9
shocks to monetary policy lending channel see credit channel
analysis 45±6 lending view 171, 175±6, 201
identi®cation procedures 31 macroeconometric analyses 288
impulse responses 20±2f, 25±6f, 28, methods of analysis 1
29±30f models 68±9, 112±15, 125, 160±1
versus systematic policy 12±15, 18±19 objective function 122, 124
VAR systems 14±15, 33, 52±3 strength 177±81, 181t
Subject Index 323
theories 170±1, 174±6

United States
traditional money view 175
banking industry 295, 296t, 312
(see also Federal Reserve System)
unemployment 265, 277, 281, 287, 299 ®nancial sector structure 311±12
United Kingdom
banking industry 219, 222 (see also
Bank of England) wage behaviour
credit channel 227±8, 228f, 241, 243f, nominal wages 260, 261, 262t, 263t,
247, 249 264, 266±7, 276
®nancial systems 219±20, 222±3, 222t, real wages 267, 268t, 269f, 269t,
224f, 225±9, 228f, 241, 243f, 250 278
interest rate channel 227±8, 228f, 242, and transmission process 287
243±5f, 246, 246t, 248 wealth effect 124±5, 127, 248
monetary policy 242, 246, 249±50

The Monetary Transmission Process Recent Developme

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Monetary Transmission Process Recent Developme

Uploaded by

Copyright:

Available Formats

The Monetary Transmission

This page intentionally left blank

The Deutsche Bundesbank

All rights reserved. No reproduction, copy or transmission of

This book is printed on paper suitable for recycling and

Includes bibliographical references and index.

ISBN 0–333–77244–X (cloth)

1. Monetary policy—Europe. 2. Finance—Europe. I. Deutsche

HG925 .M6646 2000

Printed and bound in Great Britain by

Notes on the Contributors vii

1 Analysis of the Monetary Transmission Mechanism:

Subject Index 319

lgnazio Angeloni, European Central Bank

Against this background Bennett McCallum in Chapter 1 considers

Differences in the monetary transmission process in European countries

respective ®nancial systems. Finally, he criticises some of Cecchetti's empirical

why Schmidt overstates the differences in the transmission of monetary policy

he warns of the risk of inappropriate harmonisations of labour markets ± for

Analysis of the Monetary Transmission

The purpose of this chapter is to consider several methodological issues

An outline of the chapter is as follows. In Section 2, it is argued that study of

2 Shocks versus systematic policy

A large volume of literature exists, much of it highly sophisticated, in which

estimations of monthly Bundesbank, Bank of Japan and Federal Reserve

yt  b0  b1 Rt � Et pt1   Et yt1  vt b1 < 0 1:1

Fuhrer±Moore (1995) speci®cation, but its theoretical properties are arguably

yt  10  11 yt�1  12 vt  13 et 1:4

pt     1 � =b1 1 yt�1 1:7

3 Effects of systematic policy

Here Ct is a CES consumption index, Mt /Pt is real domestic money balances, vt

inputs and the state of technology, the latter driven by an exogenous

Rt  1 � 3 Et�1 pt1  0  1 Et�1 pt �  

In (1.9), one difference relative to (1.3) is the inclusion of a lagged Rt term, to

Table 1.1 Simulation results for variants of policy rule (1.9)

Case std dev. of 1  0.1 0.5 1.0 5.0 50.0

Et�1 pt as target p 10.54 7.37 5.48 1.91 0.24

Et�1 pt1 as target p 11.21 9.76 8.57 4.51 **

2  0.1 0.5 1.0 5.0 50.0

consumption in relation to future consumption. Impulse response functions

a Responses to unit shock to IS; μ1 =

Figure 1.1 Impulse response functions, basic model (IS shock)

a Response to unit shock to UIP; μ1 = 0.1

b Response to unit shock to UIP; μ1 = 5.0

Figure 1.2 Impulse response functions; basic model (UIP shock)

Figure 1.3 Impulse response functions, basic model (y shock)

Table 1.2 Simulation results for model variants

Case std. dev. of 1  0.1 0.5 1.0 5.0 50.0

h  0.8 p 10.54 7.37 5.48 1.91 0.24

h  0.0 p 7.01 6.09 5.33 2.54 0.37

Closed econ. p 18.33 5.98 2.87 0.59 0.06

F±M price adj p 3.87 3.48 3.15 2.54 1.87

F±M price adj p 3.97 3.07 3.36 2.63 2.00

a good model, but is useful in tending to reduce inconsistencies and forcing

setting h  0.0 eliminates a source of consumption persistence that obtains

a Responses to unit shock to policy rule, with h = 0.8

b Response to unit shock to policy rule, with h = 0.0

Figure 1.4 Impulse responses to monetary shock

a Responses to unit shock to IS; h = 0.8

Figure 1.5 Impulse responses to shock to IS

Table 1.3 Autocorrelation functions

US dataa Basic model Model with (1.10) replacing (1.2)

Lag y~t pt y~t pt y~t pt

yt b0 b1 Rt � Et pt1 Et yt1 vt b1 < 0 1:1

yt 10 11 yt�1 12 vt 13 et 1:4

pt 1 � =b1 1 yt�1 1:7

Rt 1 � 3 Et�1 pt1 0 1 Et�1 pt �

Case std dev. of 1 0.1 0.5 1.0 5.0 50.0

Et�1 pt1 as target p 11.21 9.76 8.57 4.51 **

2 0.1 0.5 1.0 5.0 50.0

Case std. dev. of 1 0.1 0.5 1.0 5.0 50.0

h 0.8 p 10.54 7.37 5.48 1.91 0.24

h 0.0 p 7.01 6.09 5.33 2.54 0.37

setting h 0.0 eliminates a source of consumption persistence that obtains

pt 1 � !Et pt1 !pt�1 y~ t ut 1:10

Figure 1.10 Autocorrelation functions for basic model (h 0.0)

Figure 1.12 Autocorrelation functions for model with (1.10) (h 0.0)

Rt f past and available present data t

Vector-autoregressions (VARs) identify the monetary policy shock t in the

pt pt�1 2:3

where 0 is a constant (low or zero) in¯ation rate.4 Similarly, we can

pt pt�1 2:5

t pt � pt�1 2:6

instead of a constant . Hence, (2.5.) illustrates the base drift in in¯ation